]> ForCES Protocol Specification Zhejiang Gongshang University
149 Jiaogong Road Hangzhou 310035 P.R.China +86-571-88071024 donglg@mail.zjgsu.edu.cn
Lulea University of Technology
Rainbow Way Lulea SE-971 87 Sweden +46 73 277 1788 avri@ltu.se
Nokia
5, Wayside Road Burlington, MA 310035 USA +1-781-993-3685 ram.gopal@nsn.com
IBM
Saumerstrasse 4 8803 Ruschlikon Switzerland rha@zurich.ibm.com
Znyx
Ottawa, Ontario Canada hadi@znyx.com
Intel
2111 NE 25th Avenue Hillsboro, OR 97124 USA +1 503 264 0334 hormuzd.m.khosravi@intel.com
Zhejiang Gongshang University
149 Jiaogong Road Hangzhou 310035 P.R.China +86-571-88057712 wmwang@mail.zjgsu.edu.cn
Routing RFC Request for Comments I-D Internet-Draft ForCES Routing Protocol This document specifies the Forwarding and Control Element Separation (ForCES) protocol. ForCES protocol is used for communications between Control Elements(CEs) and Forwarding Elements (FEs) in a ForCES Network Element (ForCES NE). This specification is intended to meet the ForCES protocol requirements defined in RFC3654. Besides the ForCES protocol, this specification also defines the requirements for the Transport Mapping Layer (TML). The participants in the ForCES Protocol Team, primary co-authors and co-editors, of this protocol specification, are: Ligang Dong (Zhejiang Gongshang University), Avri Doria (Lulea University of Technology), Ram Gopal (Nokia), Robert Haas (IBM), Jamal Hadi Salim (Znyx), Hormuzd M Khosravi (Intel), and Weiming Wang (Zhejiang Gongshang University). Special acknowledgement goes to Joel Halpern who has done extensive editing in support of congruence between the model and this protocol specification. Without his participation and persistence, this specification might never have been completed.
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC 2119.
In and the following notation is used to indicate multiplicity: (value)+ .... means "1 or more instance of value" (value)* .... means "0 or more instance of value"
All integers are to be coded as unsigned binary integers of appropriate length.
Forwarding and Control Element Separation (ForCES) defines an architectural framework and associated protocols to standardize information exchange between the control plane and the forwarding plane in a ForCES Network Element (ForCES NE). RFC 3654 has defined the ForCES requirements, and RFC 3746 has defined the ForCES framework. While there may be multiple protocols used within the overall ForCES architecture, the term "ForCES protocol" and "protocol" as used in this document refers to the protocol used to standardize the information exchange between Control Elements (CEs) and Forwarding Elements (FEs) only. The ForCES FE model presents a formal way to define FE Logical Function Blocks (LFBs) using XML. LFB configuration components, capabilities, and associated events are defined when the LFB is formally created. The LFBs within the FE are accordingly controlled in a standardized way by the ForCES protocol. This document defines the ForCES protocol specifications. The ForCES protocol works in a master-slave mode in which FEs are slaves and CEs are masters. The protocol includes commands for transport of Logical Function Block (LFB) configuration information, association setup, status, and event notifications, etc. provides a glossary of terminology used in the specification. provides an overview of the protocol, including a discussion on the protocol framework, descriptions of the Protocol Layer (PL) and a Transport Mapping Layer (TML), as well as of the ForCES protocol mechanisms. describes several Protocol scenarios and includes message exchange descriptions. While this document does not define the TML, details the services that a TML MUST provide (TML requirements). The ForCES protocol defines a common header for all protocol messages. The header is defined in , while the protocol messages are defined in . describes the protocol support for high availability mechanisms including redundancy and fail over. defines the security mechanisms provided by the PL and TML.
This document follows the terminology defined by the ForCES Requirements in and by the ForCES framework in . The definitions below are repeated below for clarity. Addressable Entity (AE) - A physical device that is directly addressable given some interconnect technology. For example, on IP networks, it is a device which can be reached using an IP address; and on a switch fabric, it is a device which can be reached using a switch fabric port number. Control Element (CE) - A logical entity that implements the ForCES protocol and uses it to instruct one or more FEs on how to process packets. CEs handle functionality such as the execution of control and signaling protocols. CE Manager (CEM) - A logical entity responsible for generic CE management tasks. It is particularly used during the pre-association phase to determine with which FE(s) a CE should communicate. This process is called FE discovery and may involve the CE manager learning the capabilities of available FEs. A logical entity that operates in the pre-association phase and is responsible for determining to which FE(s) a CE should communicate. This process is called FE discovery and may involve the CE manager learning the capabilities of available FEs. A CE manager may use anything from a static configuration to a pre-association phase protocol (see below) to determine which FE(s) to use. Being a logical entity, a CE manager might be physically combined with any of the other logical entities such as CEs. Datapath - A conceptual path taken by packets within the forwarding plane inside an FE. Forwarding Element (FE) - A logical entity that implements the ForCES protocol. FEs use the underlying hardware to provide per-packet processing and handling as directed/controlled by one or more CEs via the ForCES protocol. FE Model - A model that describes the logical processing functions of an FE. The FE model is defined using Logical Function Blocks (LFBs). FE Manager (FEM) - A logical entity responsible for generic FE management tasks. It is used during pre-association phase to determine with which CE(s) an FE should communicate. This process is called CE discovery and may involve the FE manager learning the capabilities of available CEs. An FE manager may use anything from a static configuration to a pre-association phase protocol (see below) to determine which CE(s) to use. Being a logical entity, an FE manager might be physically combined with any of the other logical entities such as FEs. ForCES Network Element (NE) - An entity composed of one or more CEs and one or more FEs. To entities outside an NE, the NE represents a single point of management. Similarly, an NE usually hides its internal organization from external entities. High Touch Capability - This term will be used to apply to the capabilities found in some forwarders to take action on the contents or headers of a packet based on content other than what is found in the IP header. Examples of these capabilities include quality of service policies, virtual private networks, firewall, and L7 content recognition. Inter-FE Topology - See FE Topology. Intra-FE Topology - See LFB Topology. LFB (Logical Function Block) - The basic building block that is operated on by the ForCES protocol. The LFB is a well defined, logically separable functional block that resides in an FE and is controlled by the CE via ForCES protocol. The LFB may reside at the FE's datapath and process packets or may be purely an FE control or configuration entity that is operated on by the CE. Note that the LFB is a functionally accurate abstraction of the FE's processing capabilities, but not a hardware-accurate representation of the FE implementation. FE Topology - A representation of how the multiple FEs within a single NE are interconnected. Sometimes this is called inter-FE topology, to be distinguished from intra-FE topology (i.e., LFB topology). LFB Class and LFB Instance - LFBs are categorized by LFB Classes. An LFB Instance represents an LFB Class (or Type) existence. There may be multiple instances of the same LFB Class (or Type) in an FE. An LFB Class is represented by an LFB Class ID, and an LFB Instance is represented by an LFB Instance ID. As a result, an LFB Class ID associated with an LFB Instance ID uniquely specifies an LFB existence. LFB Metadata - Metadata is used to communicate per-packet state from one LFB to another, but is not sent across the network. The FE model defines how such metadata is identified, produced and consumed by the LFBs. It defines the functionality but not how metadata is encoded within an implementation. LFB Attribute - Operational parameters of the LFBs that must be visible to the CEs are conceptualized in the FE model as the LFB attributes. The LFB attributes include, for example, flags, single parameter arguments, complex arguments, and tables that the CE can read and/or write via the ForCES protocol (see below). LFB Topology - Representation of how the LFB instances are logically interconnected and placed along the datapath within one FE. Sometimes it is also called intra-FE topology, to be distinguished from inter-FE topology. Pre-association Phase - The period of time during which an FE Manager and a CE Manager are determining which FE(s) and CE(s) should be part of the same network element. Post-association Phase - The period of time during which an FE knows which CE is to control it and vice versa. This includes the time during which the CE and FE are establishing communication with one another. ForCES Protocol - While there may be multiple protocols used within the overall ForCES architecture, the term "ForCES protocol" and "protocol" refer to the Fp reference points in the ForCES Framework in . This protocol does not apply to CE-to-CE communication, FE-to-FE communication, or to communication between FE and CE managers. Basically, the ForCES protocol works in a master-slave mode in which FEs are slaves and CEs are masters. This document defines the specifications for this ForCES protocol. ForCES Protocol Layer (ForCES PL) - A layer in the ForCES protocol architecture that defines the ForCES protocol messages, the protocol state transfer scheme, as well as the ForCES protocol architecture itself (including requirements of ForCES TML as shown below). Specifications of ForCES PL are defined by this document. ForCES Protocol Transport Mapping Layer (ForCES TML) - A layer in ForCES protocol architecture that uses the capabilities of existing transport protocols to specifically address protocol message transportation issues, such as how the protocol messages are mapped to different transport media (like TCP, IP, ATM, Ethernet, etc), and how to achieve and implement reliability, multicast, ordering, etc. The ForCES TML specifications are detailed in separate ForCES documents, one for each TML.
The reader is referred to the Framework document , and in particular sections 3 and 4, for an architectural overview and an explanation of how the ForCES protocol fits in. There may be some content overlap between the framework document and this section in order to provide clarity. This document is authoritative on the protocol whereas is authoritative on the architecture.
below is reproduced from the Framework document for clarity. It shows a NE with two CEs and two FEs.
The ForCES protocol domain is found in the Fp Reference Points. The Protocol Element configuration reference points, Fc and Ff also play a role in the booting up of the ForCES Protocol. The protocol element configuration (indicated by reference points Fc, Ff, and Fl in ) is out of scope of the ForCES protocol but is touched on in this document in discussion of FEM and CEM since it is an integral part of the protocol pre-association phase.
below shows further breakdown of the Fp interfaces by means of the example of an MPLS QoS enabled Network Element.
The ForCES Interface shown in constitutes two pieces: the PL and the TML.
This is depicted in below.
The PL is in fact the ForCES protocol. Its semantics and message layout are defined in this document. The TML Layer is necessary to connect two ForCES PLs as shown in above. The TML is out of scope for this document but is within scope of ForCES. This document defines requirements the PL needs the TML to meet. Both the PL and the TML are standardized by the IETF. While only one PL is defined, different TMLs are expected to be standardized. To interoperate the TML at the CE and FE are expected to conform to the same definition. On transmit, the PL delivers its messages to the TML. The local TML delivers the message to the destination TML. On receive, the TML delivers the message to its destination PL.
The PL is common to all implementations of ForCES and is standardized by the IETF as defined in this document. The PL is responsible for associating an FE or CE to an NE. It is also responsible for tearing down such associations. An FE uses the PL to transmit various subscribed-to events to the CE PL as well as to respond to various status requests issued from the CE PL. The CE configures both the FE and associated LFBs' operational parameters using the PL. In addition the CE may send various requests to the FE to activate or deactivate it, reconfigure its HA parameterization, subscribe to specific events etc. More details can be found in .
The TML transports the PL messages. The TML is where the issues of how to achieve transport level reliability, congestion control, multicast, ordering, etc. are handled. It is expected that more than one TML will be standardized. The various possible TMLs could vary their implementations based on the capabilities of underlying media and transport. However, since each TML is standardized, interoperability is guaranteed as long as both endpoints support the same TML. All ForCES Protocol Layer implementations MUST be portable across all TMLs, because all TMLs MUST have the top edge semantics defined in this document.
The FEM and CEM components, although valuable in the setup and configurations of both the PL and TML, are out of scope of the ForCES protocol. The best way to think of them is as configurations/parameterizations for the PL and TML before they become active (or even at runtime based on implementation). In the simplest case, the FE or CE reads a static configuration file. RFC 3746 has a more detailed description on how the FEM and CEM could be used. The pre-association phase, where the CEM and FEM can be used, are described briefly in . An example of typical things the FEM/CEM could configure would be TML specific parameterizations such as: How the TML connection should happen (for example what IP addresses to use, transport modes etc); The ID for the FE (FEID) or CE (CEID) (which would also be issued during the pre-association phase). Security parameterization such as keys etc. Connection association parameters Example of connection association parameters this might be: simple parameters: send up to 3 association messages every 1 second complex parameters: send up to 4 association messages with increasing exponential timeout
ForCES, in relation to NEs, involves two phases: the pre-association phase, where configuration/initialization/bootup of the TML and PL layer happens, and the post-association phase where the ForCES protocol operates to manipulate the parameters of the FEs.
--->------------>---->---->---->------->----+ | Y ^ | | Y +---+-------+ +-------------+ |FE Pre- | | FE Post- | |Association| CE sends Association Teardown | Association | |Phase |<------- <------<-----<------<-------+ Phase | | | | | +-----------+ +-------------+ ^ Y | | +-<---<------<-----<------<----<---------<------+ FE loses association ]]>
In the mandated case, once associated, the FE may forward packets depending on the configuration of its specific LFBs. An FE which is associated to a CE will continue sending packets until it receives an Association teardown message or until it loses association. An unassociated FE MAY continue sending packets when it has a high availability capability. The extra details are explained in and not discussed here to allow for a clear explanation of the basics. The FE state transitions are controlled by means of the FE Object LFB FEState component, which is defined in section 5.1 and also explained in . The FE initializes in the FEState OperDisable. When the FE is ready to process packets in the data path it transitions itself to the OperEnable state. The CE may decide to pause the FE after it already came up as OperEnable. It does this by setting the FEState to AdminDisable. The FE stays in the AdminDisable state until it is explicitly configured by the CE to transition to the OperEnable state. When the FE loses its association with the CE it may go into the pre-association phase depending on the programmed policy. For the FE to properly complete the transition to the AdminDisable state, it MUST stop Packet forwarding and this may impact multiple LFBS. How this is achieved is outside the scope of this specification.
The ForCES interface is configured during the pre-association phase. In a simple setup, the configuration is static and is typically read from a saved configuration file. All the parameters for the association phase are well known after the pre-association phase is complete. A protocol such as DHCP may be used to retrieve the configuration parameters instead of reading them from a static configuration file. Note, this will still be considered static pre-association. Dynamic configuration may also happen using the Fc, Ff and Fl reference points (refer to ). Vendors may use their own proprietary service discovery protocol to pass the parameters. Essentially, only guidelines are provided here and the details are left to the implementation.
The following are scenarios reproduced from the Framework Document to show a pre-association example. <--Fc ref pt-------> FE Manager FE CE Manager CE | | | | | | | | (security exchange) (security exchange) 1|<------------>| authentication 1|<----------->|authentication | | | | (FE ID, components) (CE ID, components) 2|<-------------| request 2|<------------|request | | | | 3|------------->| response 3|------------>|response (corresponding CE ID) (corresponding FE ID) | | | | | | | | ]]>
| FE Manager FE CE Manager CE | | | | | | | | (security exchange) | | 1|<------------------------------>| | | | | | (a list of CEs and their components) | 2|<-------------------------------| | | | | | (a list of FEs and their components) | 3|------------------------------->| | | | | | | | | | ]]>
Before the transition to the association phase, the FEM will have established contact with a CEM component. Initialization of the ForCES interface will have completed, and authentication as well as capability discovery may be complete. Both the FE and CE would have the necessary information for connecting to each other for configuration, accounting, identification, and authentication purposes. To summarize, at the completion of this stage both sides have all the necessary protocol parameters such as timers, etc. The Fl reference point may continue to operate during the association phase and may be used to force a disassociation of an FE or CE. The specific interactions of the CEM and the FEM that are part of the pre-association phase are out of scope; for this reason these details are not discussed any further in this specification. The reader is referred to the framework document for a slightly more detailed discussion.
In this phase, the FE and CE components communicate with each other using the ForCES protocol (PL over TML) as defined in this document. There are three sub-phases: Association Setup Stage Established Stage Association Lost Stage
The FE attempts to join the NE. The FE may be rejected or accepted. Once granted access into the NE, capabilities exchange happens with the CE querying the FE. Once the CE has the FE capability information, the CE can offer an initial configuration (possibly to restore state) and can query certain components within either an LFB or the FE itself. More details are provided in . On successful completion of this stage, the FE joins the NE and is moved to the Established Stage.
In this stage, the FE is continuously updated or queried. The FE may also send asynchronous event notifications to the CE or synchronous heartbeat notifications if programmed to do so. This stage continues until a termination occurs, either due to loss of connectivity or due to a termination initiated by either the CE or the FE. Refer to the section on protocol scenarios, , for more details.
In this state, both or either the CE or FE declare the other side is no longer associated. The disconnection could be initiated by either party for administrative purposes but may also be driven by operational reasons such as loss of connectivity. A core LFB known as FE Protocol Object (FEPO) is defined (refer to and ). FEPO defines various timers which can be used in conjunction with traffic sensitive heartbeat mechanism () to detect loss of connectivity. The loss of connectivity between TMLs does not indicate a loss of association between respective PL layers. If the TML cannot repair the transport loss before the programmed FEPO timer thresholds associated with the FE is exceeded, then the association between the respective PL layers will be lost. FEPO defines several policies that can be programmed to define behavior upon a detected loss of association. The FEPO's programmed CE failover policy (refer to , , and ) defines what takes place upon loss of association. For this version of the protocol (as defined in this document), the FE, upon re-association, MUST discard any state it has as invalid and retrieve new state. This approach is motivated by a desire for simplicity (as opposed to efficiency).
Various semantics are exposed to the protocol users via the PL header including: transaction capabilities, atomicity of transactions, two phase commits, batching/parallelization, high availability and failover as well as command pipelines. The EM (Execute Mode) flag, AT (Atomic Transaction) flag, and TP (Transaction Phase) flag as defined in the Common Header () are relevant to these mechanisms.
In the master-slave relationship the CE instructs one or more FEs on how to execute operations and how to report the results. This section details the different modes of execution that a CE can order the FE(s) to perform, as defined in . It also describes the different modes a CE can ask the FE(s) to use for formatting the responses after processing the operations as requested. These modes relate to the transactional two phase commitment operations.
There are 3 execution modes that can be requested for a batch of operations spanning one or more LFB selectors (refer to ) in one protocol message. The EM flag defined in the Common Header selects the execution mode for a protocol message, as below: execute-all-or-none continue-execute-on-failure execute-until-failure
When set to this mode of execution, independent operations in a message MAY be targeted at one or more LFB selectors within an FE. All these operations are executed serially and the FE MUST have no execution failure for any of the operations. If any operation fails to execute, then all the previous ones that have been executed prior to the failure will need to be undone. I.e., there is rollback for this mode of operation. Refer to for how this mode is used in cases of transactions. In such a case, no operation is executed unless a commit is issued by the CE. Care should be taken on how this mode is used because a mis-configuration could result in traffic losses. To add certainty to the success of an operation, one should use this mode in a transactional operation as described in
If several independent operations are targeted at one or more LFB selectors, execution continues for all operations at the FE even if one or more operations fail.
In this mode all operations are executed on the FE sequentially until the first failure. The rest of the operations are not executed but operations already completed are not undone. I.e., there is no rollback in this mode of operation.
A transaction is defined as a collection of one or more ForCES operations within one or more PL messages that MUST meet the ACIDity properties , defined as: In a transaction involving two or more discrete pieces of information, either all of the pieces are committed or none are. A transaction either creates a new and valid state of data, or, if any failure occurs, returns all data to the state it was in before the transaction was started. A transaction in process and not yet committed MUST remain isolated from any other transaction. Committed data is saved by the system such that, even in the event of a failure and a system restart, the data is available in its correct state. There are cases where the CE knows exact memory and implementation details of the FE such as in the case of an FE-CE pair from the same vendor where the FE-CE pair is tightly coupled. In such a case, the transactional operations may be simplified further by extra computation at the CE. This view is not discussed further other than to mention that it is not disallowed. As defined above, a transaction is always atomic and MAY be Within an FE alone Example: updating multiple tables that are dependent on each other. If updating one fails, then any that were already updated MUST be undone. Distributed across the NE Example: updating table(s) that are inter-dependent across several FEs (such as L3 forwarding related tables).
By use of the execute mode, as defined in , the protocol has provided a mechanism for transactional operations within one stand-alone message. The 'execute-all-or-none' mode can meet the ACID requirements. For transactional operations of multiple messages within one FE or across FEs, a classical transactional protocol known as Two Phase Commit (2PC) is supported by the protocol to achieve the transactional operations utilizing Config messages (). The COMMIT and TRCOMP operations in conjunction with the AT and the TP flags in Common Header are provided for 2PC-based transactional operations spanning multiple messages. The AT flag, when set, indicates this message belongs to an Atomic Transaction. All messages for a transaction operation MUST have the AT flag set. If not set, it means the message is a stand-alone message and does not participate in any transaction operation that spans multiple messages. The TP flag indicates the Transaction Phase this message belongs to. There are four (4) possible phases for an transactional operation known as: SOT (Start of Transaction) MOT (Middle of Transaction) EOT (End of Transaction) ABT (Abort) The COMMIT operation is used by the CE to signal to the FE(s) to commit a transaction. When used with an ABT TP flag, the COMMIT operation signals the FE(s) to rollback (i.e un-COMMIT) a previously committed transaction. The TRCOMP operation is a small addition to the classical 2PC approach. TRCOMP is sent by the CE to signal the FE(s) that the transaction they have COMMITed is now over. This allows the FE(s) an opportunity to clear state they may have kept around to perform a rollback (if it became necessary). A transaction operation is started with a message in which the TP flag is set to Start of Transaction (SOT). Multi-part messages, after the first one, are indicated by the Middle of Transaction flag (MOT). All messages from the CE MUST set the AlwaysACK flag () to solicit responses from the FE(s). Before the CE issues a commit (described further below) the FE MUST only validate that the operation can be executed but not execute it. Any failure notified by an FE causes the CE to abort the transaction on all FEs involved in the transaction. This is achieved by sending a Config message with an ABT flag and a COMMIT operation. If there are no failures by any participating FE, the transaction commitment phase is signaled from the CE to the FE by an End of Transaction (EOT) configuration message with a COMMIT operation. The FE MUST respond to the CE's EOT message. There are two possible failure scenarios in which the CE MUST abort the transaction (as described above): If any participating FE responds with a failure message in relation to the transaction. If no response is received from a participating FE within a specified timeout. If all participating FE(s) respond with a success indicator within the expected time, then the CE MUST issue a TRCOMP operation to all participating FEs. An FE MUST NOT respond to a TRCOMP. Note that a transactional operation is generically atomic, therefore it requires that the execute modes of all messages in a transaction operation should always be kept the same and be set to 'execute-all-or-none'. If the EM flag is set to other execute modes, it will result in a transaction failure. As noted above, a transaction may span multiple messages. It is up to the CE to keep track of the different outstanding messages making up a transaction. As an example, the correlator field could be used to mark transactions and a sequence field to label the different messages within the same atomic transaction, but this is out of scope and up to implementations.
Any of the participating FEs, or the CE, or the associations between them, may fail after the EOT response message has been sent by the FE but before the CE has received all the responses, e.g. if the EOT response never reaches the CE. In this protocol revision, as indicated in , an FE losing an association would be required to get entirely new state from the newly associated CE upon a re-association. Although this approach is simplistic and provides likeliness of loosing datapath traffic, it is a design choice to avoid the additional complexity of managing graceful restarts. The HA mechanisms () are provided to allow for a continuous operation in case of FE failures. Flexibility is provided on how to react when an FE looses association. This is dictated by the CE Failover policy (refer to and ).
This section illustrates an example of how a successful two phase commit between a CE and an FE would look like in the simple case.
| | | | (3) Config, MOT,AT, EM=All-or-None, OP= SET/DEL,etc | |<-----------------------------------------------------| | | | (4) ACKnowledge | |----------------------------------------------------->| | | | (5) Config, MOT,AT, EM=All-or-None, OP= SET/DEL,etc | |<-----------------------------------------------------| | | | (6) ACKnowledge | |----------------------------------------------------->| . . . . . . . . | | | (N) Config, EOT,AT, EM=All-or-None, OP= COMMIT | |<-----------------------------------------------------| | | | (N+1)Config-response, ACKnowledge, OP=COMMIT-RESPONSE| |----------------------------------------------------->| | | | (N+2) Config, OP=TRCOMP | |<-----------------------------------------------------| ]]>
For the scenario illustrated above: In step #1, the CE issues a Config message with an operation of choice like a DEL or SET. The transactional flag are set to indicate a Start of Transaction (SOT), Atomic Transaction (AT), execute-all-or-none. The FE validates that it can execute the request successfully and then issues an acknowledgment back to the the CE in step #2. In step #3, the same sort of construct as in step #1 is repeated by the CE with the transaction flag changed to Middle of Transaction(MOT). The FE validates that it can execute the request successfully and then issues an acknowledgment back to the the CE in step #4. The CE-FE exchange continues in the same manner until all the operations and their parameters are transferred to the FE. This happens in step #(N-1). In step #N, the CE issues a commit. A commit is a config message with an operation of type COMMIT. The transactional flags are set to End of Transaction (EOT). Essentially, this is an "empty" message asking the FE to execute all the operations it has gathered since the beginning of the transaction (message #1). The FE at this point executes the full transaction. It then issues an acknowledgment back to the the CE in step #(N+1) which contains a COMMIT-RESPONSE. The CE in this case has the simple task of issuing a TRCOMP operation the the FE in step #(N+2)
It is desirable that the PL not become the bottleneck when larger bandwidth pipes become available. To pick a hypothetical example in today's terms, if a 100Gbps pipe is available and there is sufficient work then the PL should be able to take advantage of this and use all of the 100Gbps pipe. Two mechanisms have been provided to achieve this. The first one is batching and the second one is a command pipeline. Batching is the ability to send multiple commands (such as Config) in one Protocol Data Unit (PDU). The size of the batch will be affected by, amongst other things, the path MTU. The commands may be part of the same transaction or may be part of unrelated transactions that are independent of each other. Command pipelining allows for pipelining of independent transactions which do not affect each other. Each independent transaction could consist of one or more batches.
There are several batching levels at different protocol hierarchies. multiple PL PDUs can be aggregated under one TML message multiple LFB classes and instances (as indicated in the LFB selector) can be addressed within one PL PDU Multiple operations can be addressed to a single LFB class and instance
The protocol allows any number of messages to be issued by the CE before the corresponding acknowledgments (if requested) have been returned by the FE. Hence pipelining is inherently supported by the protocol. Matching responses with requests messages can be done using the correlator field in the message header.
Heartbeats (HB) between FEs and CEs are traffic sensitive. An HB is sent only if no PL traffic is sent between the CE and FE within a configured interval. This has the effect of reducing the amount of HB traffic in the case of busy PL periods. An HB can be sourced by either the CE or FE. When sourced by the CE, a response can be requested (similar to the ICMP ping protocol). The FE can only generate HBs in the case of being configured to do so by the CE. Refer to and for details.
All PL messages operate on LFB constructs, as this provides more flexibility for future enhancements. This means that maintenance and configurability of FEs, NE, as well as the ForCES protocol itself MUST be expressed in terms of this LFB architecture. For this reason special LFBs are created to accommodate this need. In addition, this shows how the ForCES protocol itself can be controlled by the very same type of structures (LFBs) it uses to control functions such as IP forwarding, filtering, etc. To achieve this, the following specialized LFBs are introduced: FE Protocol LFB which is used to control the ForCES protocol. FE Object LFB which is used to control attributes relative to the FE itself. Such attributes include FEState , vendor, etc. These LFBs are detailed in .
This section provides a very high level description of sample message sequences between a CE and FE. For protocol message encoding refer to and for the semantics of the protocol refer to .
The associations among CEs and FEs are initiated via Association setup message from the FE. If a setup request is granted by the CE, a successful setup response message is sent to the FE. If CEs and FEs are operating in an insecure environment then the security associations have to be established between them before any association messages can be exchanged. The TML MUST take care of establishing any security associations. This is typically followed by capability query, topology query, etc. When the FE is ready to start processing the data path, it sets the FEO FEState component to OperEnable (Refer to for details) and reports it to the CE as such when it is first queried. Typically the FE is expected to be ready to process the data path before it associates, but there maybe rare cases where it needs time do some pre-processing - in such a case the FE will start in the OperDisable state and when it is ready will transition to OperEnable state. An example of an FE starting in the OperDisable then transitioning to OperEnable is illustrated in . The CE could at any time also disable the FEs datapath operations by setting the FEState to AdminDisable. The FE MUST NOT process packets during this state unless the CE sets the state back to OperEnable. These sequences of messages are illustrated in below.
| | | | Asso Setup Resp | |<----------------------| | | | LFBx Query capability | |<----------------------| | | | LFBx Query Resp | |---------------------->| | | | FEO Query (Topology) | |<----------------------| | | | FEO Query Resp | |---------------------->| | | | FEO OperEnable Event | |---------------------->| | | | Config FEO Adminup | |<----------------------| | | | FEO Config-Resp | |---------------------->| | | ]]>
On successful completion of this state, the FE joins the NE.
In this state, the FE is continuously updated or queried. The FE may also send asynchronous event notifications to the CE, synchronous heartbeat messages, or packet redirect message to the CE. This continues until a termination (or deactivation) is initiated by either the CE or FE. below, helps illustrate this state.
| | | | Heart Beat | |----------------------------->| | | | Config-set LFBy (Event sub.) | |<-----------------------------| | | | Config Resp LFBy | |----------------------------->| | | | Config-set LFBx Attr | |<-----------------------------| | | | Config Resp LFBx | |----------------------------->| | | |Config-Query LFBz (Stats) | |<--------------------------- -| | | | Query Resp LFBz | |----------------------------->| | | | FE Event Report | |----------------------------->| | | | Config-Del LFBx Attr | |<-----------------------------| | | | Config Resp LFBx | |----------------------------->| | | | Packet Redirect LFBx | |----------------------------->| | | | Heart Beat | |<-----------------------------| . . . . | | ]]>
Note that the sequence of messages shown in the figure serve only as examples and the message exchange sequences could be different from what is shown in the figure. Also, note that the protocol scenarios described in this section do not include all the different message exchanges that would take place during failover. That is described in the HA section () .
The requirements below are expected to be met by the TML. This text does not define how such mechanisms are delivered. As an example, the mechanisms to meet the requirements could be defined to be delivered via hardware or between 2 or more TML software processes on different CEs or FEs in protocol level schemes. Each TML MUST describe how it contributes to achieving the listed ForCES requirements. If for any reason a TML does not provide a service listed below a justification needs to be provided. Implementations that support the ForCES Protocol Specification MUST IMPLEMENT . Note that additional TMLs might be specified in the future, and if a new TML defined in the future that meets the requirements listed here proves to be better, then the MUST IMPLEMENT TML may redefined. Reliability Various ForCES messages will require varying degrees of reliable delivery via the TML. It is the TML's responsibility to provide these shades of reliability and describe how the different ForCES messages map to reliability. The most common level of reliability is what we refer to as strict or robust reliability in which we mean: no losses, corruption, or re-ordering of information being transported while ensuring message delivery in a timely fashion. Payloads such as configuration from a CE and its response from an FE are mission critical and must be delivered in a robust reliable fashion. Thus, for information of this sort, the TML MUST either provide built-in protocol mechanisms or use a reliable transport protocol for achieving robust/strict reliability. Some information or payloads, such as redirected packets or packet sampling, may not require robust reliability (can tolerate some degree of losses). For information of this sort, the TML could define to use a mechanism that is not strictly reliable (while conforming to other TML requirements such as congestion control). Some information or payloads, such as heartbeat packets that may prefer timeliness over reliable delivery. For information of this sort, the TML could define to use a mechanism that is not strictly reliable (while conforming to other TML requirements such as congestion control). Security TML provides security services to the ForCES PL. Because a ForCES PL is used to operate an NE, attacks designed to confuse, disable, or take information from a ForCES-based NE may be seen as a prime objective during a network attack. An attacker in a position to inject false messages into a PL stream can either affect the FE's treatment of the data path (example by falsifying control data reported as coming from the CE), or the CE itself (by modifying events or responses reported as coming from the FE); for this reason, CE and FE node authentication and TML Message authentication is important. The PL messages may also contain information of value to an attacker, including information about the configuration of the network, encryption keys and other sensitive control data, so care must be taken to confine their visibility to authorized user The TML MUST provide a mechanism to authenticate ForCES CEs and FEs in order to prevent the participation of unauthorized CEs and unauthorized FEs in the control and data path processing of a ForCES NE. The TML SHOULD provide a mechanism to ensure message authentication of PL data transferred from the CE to FE (and vice-versa) in order to prevent the injection of incorrect data into PL messages. The TML SHOULD provide a mechanism to ensure the confidentiality of data transferred from the ForCES PL, in order to prevent disclosure of PL level information transported via the TML. The TML SHOULD provide these services by employing TLS or IPSEC. Congestion control The transport congestion control scheme used by the TML needs to be defined. The congestion control mechanism defined by the TML MUST prevent transport congestive collapse on either the FE or CE side. Uni/multi/broadcast addressing/delivery, if any If there is any mapping between PL and TML level uni/multi/broadcast addressing it needs to be defined. HA decisions It is expected that availability of transport links is the TML's responsibility. However, based upon its configuration, the PL may wish to participate in link failover schemes and therefore the TML MUST support this capability. Please refer to for details. Encapsulations used Different types of TMLs will encapsulate the PL messages on different types of headers. The TML needs to specify the encapsulation used. Prioritization It is expected that the TML will be able to handle up to 8 priority levels needed by the PL and will provide preferential treatment. While the TML needs to define how this is achieved, it should be noted that the requirement for supporting up to 8 priority levels does not mean that the underlying TML MUST be capable of providing up to 8 actual priority levels. In the event that the underlying TML layer does not have support for 8 priority levels, the supported priority levels should be divided between the available TML priority levels. For example, if the TML only supports 2 priority levels, the 0-3 could go in one TML priority level, while 4-7 could go in the other. The TML MUST NOT reorder config packets with the same priority. Node Overload Prevention The TML MUST define mechanisms it uses to help prevent node overload. Overload results in starvation of node compute cycles and/or bandwidth resources which reduces the operational capacity of a ForCES NE. NE node overload could be deliberately instigated by a hostile node to attack a ForCES NE and create a denial of service. It could also be created by a variety of other reasons such as large control protocol updates (eg BGP flaps) which consequently cause a high frequency of CE to FE table updates, HA failovers or component failures which migrate an FE or CE load overwhelming the new CE or FE, etc. Although the environment under which SIP and ForCES operate are different, provides a good guide to generic node requirements one needs to guard for. A ForCES node CPU may be overwhelmed because the incoming packet rate is higher than it can keep up with - in such a case a nodes transport queues grow and transport congestion subsequently follows. A ForCES node CPU may also be adversely overloaded with very few packets i.e no transport congestion at all (eg a in a DOS attack against a table hashing algorithmn which overflows the table and/or keeps the CPU busy so it does not process other tasks) The TML node-overload solution specified MUST address both types of node overload scenarios.
It is expected that it should be possible to use a configuration reference point, such as the FEM or the CEM, to configure the TML. Some of the configured parameters may include: PL ID Connection Type and associated data. For example if a TML uses IP/TCP/UDP, then parameters such as TCP and UDP port and IP addresses need to be configured. Number of transport connections Connection capability, such as bandwidth, etc. Allowed/supported connection QoS policy (or congestion control policy)
All PL PDUs start with a common header [] followed by a one or more TLVs [] which may nest other TLVs []. All fields are in network byte order.
The message is 32 bit aligned. Version number. Current version is 1. Unused at this point. A receiver should not interpret this field. Senders MUST set it to zero and receivers MUST ignore this field. Commands are defined in . length of header + the rest of the message in DWORDS (4 byte increments). Each of the source and destination IDs are 32 bit IDs which are unique NE-wide and which identify the termination points of a ForCES PL message. IDs allow multi/broad/unicast addressing with the following approach: A split address space is used to distinguish FEs from CEs. Even though in a large NE there are typically two or more orders of magnitude more FEs than CEs, the address space is split uniformly for simplicity. The address space allows up to 2^30 (over a billion) CEs and the same amount of FEs.
The 2 most significant bits called Type Switch (TS) are used to split the ID space as follows:
Multicast or broadcast IDs are used to group endpoints (such as CEs and FES). As an example one could group FEs in some functional group, by assigning a multicast ID. Likewise, subgroups of CEs that act, for instance, in a back-up mode may be assigned a multicast ID to hide them from the FE. Multicast IDs can be used for both source or destination IDs. Broadcast IDs can be used only for destination IDs. This document does not discuss how a particular multicast ID is associated to a given group though it could be done via configuration process. The list of IDs an FE owns or is part of are listed on the FE Object LFB. This field is set by the CE to correlate ForCES Request Messages with the corresponding Response messages from the FE. Essentially it is a cookie. The correlator is handled transparently by the FE, i.e., for a particular Request message the FE MUST assign the same correlator value in the corresponding Response message. In the case where the message from the CE does not elicit a response, this field may not be useful. The correlator field could be used in many implementations specific ways by the CE. For example, the CE could split the correlator into a 32-bit transactional identifier and 32-bit message sequence identifier. Another example is a 64-bit pointer to a context block. All such implementation specific use of the correlator is outside the scope of this specification. It should be noted that the correlator is transmitted on the network as if it were a 64 bit unsigned integer with the leftmost or most significant octet (bits 63-56) transmitted first. Whenever the correlator field is not relevant, because no message is expected, the correlator field is set to 0. Identified so far:
The ACK indicator flag is only used by the CE when sending a Config Message () or a HB message () to indicate to the message receiver whether or not a response is required by the sender. Note that for all other messages than the Config Message or the HB Message this flag MUST be ignored. The flag values are defined as below: 'NoACK' (0b00) - to indicate that the message receiver MUST NOT send any response message back to this message sender. 'SuccessACK'(0b01) - to indicate the message receiver MUST send a response message back only when the message has been successfully processed by the receiver. 'FailureACK'(0b10) - to indicate the message receiver MUST send a response message back only when there is failure by the receiver in processing (executing) the message. In other words, if the message can be processed successfully, the sender will not expect any response from the receiver. 'AlwaysACK' (0b11) - to indicate the message receiver MUST send a response message. Note that in above definitions, the term success implies a complete execution without any failure of the message. Anything else than a complete successful execution is defined as a failure for the message processing. As a result, for the execution modes (defined in ) like execute-all-or-none, execute-until-failure, and continue-execute-on-failure, if any single operation among several operations in the same message fails, it will be treated as a failure and result in a response if the ACK indicator has been set to 'FailureACK' or 'AlwaysACK'. Also note that, other than in Config and HB Messages, requirements for responses of messages are all given in a default way rather than by ACK flags. The default requirements of these messages and the expected responses are summarized below. Detailed descriptions can be found in the individual message definitions: Association Setup Message always expects a response. Association Teardown Message, and Packet Redirect Message, never expect responses. Query Message always expects a response. Response message never expects further responses. ForCES protocol defines 8 different levels of priority (0-7). The priority level can be used to distinguish between different protocol message types as well as between the same message type. The higher the priority value, the more important the PDU is. For example, the REDIRECT packet message could have different priorities to distinguish between routing protocols packets and ARP packets being redirected from FE to CE. The Normal priority level is 1. The different priorities imply messages could be re-ordered; however, re-ordering is undesirable when it comes to a set of messages within a transaction and caution should be exercised to avoid this from happening. There are 3 execution modes refer to for details. Reserved..................... (0b00) `execute-all-or-none` ....... (0b01) `execute-until-failure` ..... (0b10) `continue-execute-on-failure` (0b11) This flag indicates if the message is stand-alone message or one of multiple messages that belongs to 2PC transaction operations. See for details. Stand-alone message ......... (0b0) 2PC transaction message ..... (0b1) A message from the CE to the FE within a transaction could be indicative of the different phases the transaction is in. Refer to for details. SOT (start of transaction) ..... (0b00) MOT (Middle of transaction) .... (0b01) EOT (end of transaction) ........(0b10) ABT (abort) .....................(0b11)
TLVs are extensively used by the ForCES protocol. TLVs have some very nice properties which make them a good candidate for encoding the XML definitions of the LFB class model. These are: Providing for binary type-value encoding that is close to the XML string tag-value scheme. Allowing for fast generalized binary-parsing functions. Allowing for forward and backward tag compatibility. This is equivalent to the XML approach i.e old applications can ignore new TLVs and newer applications can ignore older TLVs.
The TLV type field is two octets, and semantically indicates the type of data encapsulated within the TLV. The TLV length field is two octets, and includes the length of the TLV type (2 octets), TLV Length (2 octets), and the length of the TLV data found in the value field, in octets. Note that this length is the actual length of the value, before any padding for alignment is added. The TLV value field carries the data. For extensibility, the TLV value may in fact be a TLV. Padding is required when the length is not a multiple of 32 bits, and is the minimum number of octets required to bring the TLV to a multiple of 32 bits. The length of the value before padding is indicated by the TLV Length field. Note: The value field could be empty which implies the minimal length a TLV could be is 4 (length of "T" field and length of "L" field).
TLV values can be other TLVs. This provides the benefits of protocol flexibility (being able to add new extensions by introducing new TLVs when needed). The nesting feature also allows for a conceptual optimization with the XML LFB definitions to binary PL representation (represented by nested TLVs).
The "Type" values in the TLV are global in scope. This means that wherever TLVs occur in the PDU, a specific Type value refers to the same Type of TLV. This is a design choice that was made to ease debugging of the protocol.
A slight variation of the TLV known as the ILV. This sets the type ("T") to be a 32-bit local index that refers to a ForCES component ID. (refer to ). The ILV length field is a 4 octet integer, and includes the length of the ILV type (4 octets), ILV Length (4 octets), and the length of the ILV data found in the value field, in octets. Note that, as in the case of the TLV, this length is the actual length of the value, before any padding for alignment is added.
It should be noted that the "I" values are of local scope and are defined by the data declarations from the LFB definition. Refer to for discussions on usage of ILVs.
In this section, we review a few encapsulation concepts that are used by the ForCES protocol for its operations. We briefly re-introduce two concepts, Paths and Keys, from the model draft in order to provide context. The reader is referred to for a lot of the finer details. For readability reasons, we introduce the encapsulation schemes that are used to carry content in a protocol message, namely FULLDATA-TLV, SPARSEDATA-TLV, and RESULT-TLV.
The model draft defines an XML-based language that allows for a formal definition of LFBs. This is similar to the relationship between ASN.1 and SNMP MIB definition (MIB being analogous to the LFB and ASN.1 being analogous to the XML model language). Any entity that the CE configures on an FE MUST be formally defined in an LFB. These entities could be scalars (e.g., a 32-bit IPv4 address) or vectors (such as a nexthop table). Each entity within the LFB is given a numeric 32-bit identifier known as an "component id". This scheme allows the attribute to be "addressed" in a protocol construct. These addressable entities could be hierarchical (e.g., a table column or a cell within a table row). In order to address hierarchical data, the concept of a path is introduced by the model . A path is a series of 32-bit component IDs which are typically presented in a dot-notation (e.g., 1.2.3.4). Section () formally defines how paths are used to reference data that is being encapsulated within a protocol message.
The model draft defines two ways to address table rows. The standard/common mechanism is to allow table rows to be referenced by a 32-bit index. The secondary mechanism is via Keys which allow for content addressing. An example key is a multi-field content key that uses the IP address and prefix length to uniquely reference an IPv4 routing table row. In essence, while the common scheme to address a table row is via its table index, a table row's path could be derived from a key. The KEYINFO-TLV () is used to carry the data that is used to do the lookup.
Data from or to the FE is carried in two types of TLVs: FULLDATA-TLV and SPARSEDATA-TLV. Responses to operations executed by the FE are carried in RESULT-TLVs. In FULLDATA-TLV, the data is encoded in such a way that a receiver of such data, by virtue of being armed with knowledge of the path and the LFB definition, can infer or correlate the TLV "Value" contents. This is essentially an optimization that helps reduce the amount of description required for the transported data in the protocol grammar. Refer to for an example of FULLDATA-TLVs. A number of operations in ForCES will need to reference optional data within larger structures. The SPARSEDATA-TLV encoding is provided to make it easier to encapsulate optionally appearing data components. Refer to for an example of SPARSEDATA-TLV. RESULT-TLVs carry responses back from the FE based on a config issued by the CE. Refer to for examples of RESULT-TLVs and for layout.
and discuss how to target an entity within an LFB. However, the addressing mechanism used requires that an LFB type and instance is selected first. The LFB Selector is used to select an LFB type and instance being targeted. Section () goes into more details; for our purpose, we illustrate this concept using below. More examples of layouts can be found reading further into the text (Example: ).
A protocol layer PDU consists of a Common Header (defined in Section ) and a message body. The Common Header is followed by a message-type-specific message body. Each message body is formed from one or more top-level TLVs. A top-level TLV may contain one or more sub-TLVs; these sub-TLVs are described in this document as OPER-TLVs, because they describe an operation to be done. Message Name Top-Level TLV OPER-TLV(s) Reference Association Setup (LFBselect)* REPORT Association Setup Response ASRresult-TLV none Association Teardown ASTreason-TLV none Config (LFBselect)+ (SET | SET-PROP | DEL | COMMIT | TRCOMP)+ Config Response (LFBselect)+ (SET-RESPONSE | SET-PROP-RESPONSE | DEL-RESPONSE | COMMIT-RESPONSE)+ Query (LFBselect)+ (GET | GET-PROP)+ Query Response (LFBselect)+ (GET-RESPONSE | GET-PROP-RESPONSE)+ Event Notifi- cation LFBselect REPORT Packet Redirect REDIRECT-TLV none Heartbeat none none The different messages are illustrated in . The different message type numerical values are defined in . All the TLVs values are defined in . An LFBselect TLV (refer to ) contains the LFB Classid and LFB Instance being referenced as well as the OPER-TLV(s) being applied to that reference. Each type of OPER-TLV is constrained as to how it describes the paths and selectors of interest. The following BNF describes the basic structure of an OPER-TLV and gives the details for each type
PATH-DATA-TLV identifies the exact component targeted and may have zero or more paths associated with it. The last PATH-DATA-TLV in the case of nesting of paths via the DATA construct in the case of SET, SET-PROP requests and GET-RESPONSE/GET-PROP-RESPONSE is terminated by encoded data or response in the form of either FULLDATA-TLV or SPARSEDATA-TLV or RESULT-TLV. PATH provides the path to the data being referenced. flags (16 bits) are used to further refine the operation to be applied on the Path. More on these later. IDcount(16 bit): count of 32 bit IDs IDs: zero or more 32bit IDs (whose count is given by IDcount) defining the main path. Depending on the flags, IDs could be field IDs only or a mix of field and dynamic IDs. Zero is used for the special case of using the entirety of the containing context as the result of the path. SELECTOR is an optional construct that further defines the PATH. Currently, the only defined selector is the KEYINFO-TLV, used for selecting an array entry by the value of a key field. The presence of a SELECTOR is correct only when the flags also indicate its presence. A mismatch is a protocol format error. A KEYINFO-TLV contains information used in content keying. A KeyID is used in a KEYINFO-TLV. It indicates which key for the current array is being used as the content key for array entry selection. The key's data is the data to look for in the array, in the fields identified by the key field. The information is encoded according to the rules for the contents of a FULLDATA-TLV, and represent the field or fields which make up the key identified by the KeyID. DATA may contain a FULLDATA-TLV, SPARSEDATA-TLV, a RESULT-TLV or 1 or more further PATH-DATA selection. FULLDATA-TLV and SPARSEDATA-TLV are only allowed on SET or SET-PROP requests, or on responses which return content information (GET-RESPONSE for example). PATH-DATA may be included to extend the path on any request. Note: Nested PATH-DATA-TLVs are supported as an efficiency measure to permit common subexpression extraction. FULLDATA-TLV and SPARSEDATA-TLV contain "the data" whose path has been selected by the PATH. Refer to for details. The following table summarizes the applicability and restrictions of the FULL/SPARSEDATA-TLVs and the RESULT-TLV to the OPER-TLVs. OPER-TLV DATA TLV RESULT-TLV SET (FULLDATA-TLV | SPARSEDATA-TLV)+ none SET-PROP (FULLDATA-TLV | SPARSEDATA-TLV)+ none SET-RESPONSE none (RESULT-TLV)+ SET-PROP-RESPONSE none (RESULT-TLV)+ DEL (FULLDATA-TLV | SPARSEDATA-TLV)+ none DEL-RESPONSE none (RESULT-TLV)+ GET none none GET-PROP none none GET-RESPONSE (FULLDATA-TLV)+ (RESULT-TLV)* GET-PROP-RESPONSE (FULLDATA-TLV)+ (RESULT-TLV)* REPORT (FULLDATA-TLV)+ none COMMIT none none COMMIT-RESPONSE none (RESULT-TLV)+ TRCOMP none none RESULT-TLV contains the indication of whether the individual SET or SET-PROP succeeded. RESULT-TLV is included on the assumption that individual parts of a SET request can succeed or fail separately. In summary this approach has the following characteristic: There can be one or more LFB class ID and instance ID combination targeted in a message (batch) There can one or more operations on an addressed LFB class ID/instance ID combination (batch) There can be one or more path targets per operation (batch) Paths may have zero or more data values associated (flexibility and operation specific) It should be noted that the above is optimized for the case of a single LFB class ID and instance ID targeting. To target multiple instances within the same class, multiple LFBselects are needed.
discusses the two types of DATA encodings (FULLDATA-TLV and SPARSEDATA-TLV) and the justifications for their existence. In this section we explain how they are encoded.
The scheme for encoding data used in this doc adheres to the following rules: The Value ("V" of TLV) of FULLDATA-TLV will contain the data being transported. This data will be as was described in the LFB definition. Variable sized data within a FULLDATA-TLV will be encapsulated inside another FULLDATA-TLV inside the V of the outer TLV. For example of such a setup refer to and In the case of FULLDATA-TLVs: When a table is referred to in the PATH (IDs) of a PATH-DATA-TLV, then the FULLDATA-TLV's "V" will contain that table's row content prefixed by its 32 bit index/subscript. On the other hand, when PATH flags are 00, the PATH may contain an index pointing to a row in table; in such a case, the FULLDATA-TLV's "V" will only contain the content with the index in order to avoid ambiguity.
The following flags are currently defined: SELECTOR Bit: F_SELKEY indicates that a KEY Selector is present following this path information, and should be considered in evaluating the path.
Global flags, such as the execution mode and the atomicity indicators defined in the header, apply to all operations in a message. Global flags provide semantics that are orthogonal to those provided by the operational flags, such as the flags defined in Path Data. The scope of operational flags is restricted to the operation.
The KEYINFO-TLV describes the KEY as well as associated KEY data. KEYs, used for content searches, are restricted and described in the LFB definition.
The LFBselect TLV is an instance of a TLV as defined in . The definition is as below:
The type of the TLV is "LFBselect" Length of the TLV including the T and L fields, in octets. This field uniquely recognizes the LFB class/type. This field uniquely identifies the LFB instance. It describes an operation nested in the LFBselect TLV. Note that usually there SHOULD be at least one OPER-TLV present for an LFB select TLV, but for the Association Setup Message defined in where the OPER-TLV is optional.
The OPER-TLV is a place holder in the grammar for TLVs that define operations. The different types are defined in , below. OPER-TLV TLV "Type" Comments SET 0x0001 From CE to FE. Used to create or add or update attributes SET-PROP 0x0002 From CE to FE. Used to create or add or update attribute properties SET-RESPONSE 0x0003 From FE to CE. Used to carry response of a SET SET-PROP-RESPONSE 0x0004 From FE to CE. Used to carry response of a SET-PROP DEL 0x0005 From CE to FE. Used to delete or remove an attribute DEL-RESPONSE 0x0006 From FE to CE. Used to carry response of a DEL GET 0x0007 From CE to FE. Used to retrieve an attribute GET-PROP 0x0008 From CE to FE. Used to retrieve an attribute property GET-RESPONSE 0x0009 From FE to CE. Used to carry response of a GET GET-PROP-RESPONSE 0x000A From FE to CE. Used to carry response of a GET-PROP REPORT 0x000B From FE to CE. Used to carry an asynchronous event COMMIT 0x000C From CE to FE. Used to issue a commit in a 2PC transaction COMMIT-RESPONSE 0x000D From an FE to CE. Used to confirm a commit in a 2PC transaction TRCOMP 0x000E From CE to FE. Used to indicate NE-wide success of 2PC transaction Different messages define OPER-TLV are valid and how they are used (refer to and ). SET, SET-PROP, and GET/GET-PROP requests are issued by the CE and do not carry RESULT-TLVs. On the other hand, SET-RESPONSE, SET-PROP-RESPONSE and GET-RESPONSE/GET-PROP-RESPONSE carry RESULT-TLVs. A GET-RESPONSE in response to a successful GET will have FULLDATA-TLVs added to the leaf paths to carry the requested data. For GET operations that fail, instead of the FULLDATA-TLV there will be a RESULT-TLV. For a SET-RESPONSE/SET-PROP-RESPONSE, each FULLDATA-TLV or SPARSEDATA-TLV in the original request will be replaced with a RESULT-TLV in the response. If the request set the FailureACK flag, then only those items which failed will appear in the response. If the request was for AlwaysACK, then all components of the request will appear in the response with RESULT-TLVs. Note that if a SET/SET-PROP request with a structure in a FULLDATA-TLV is issued, and some field in the structure is invalid, the FE will not attempt to indicate which field was invalid, but rather will indicate that the operation failed. Note further that if there are multiple errors in a single leaf PATH-DATA/FULLDATA-TLB, the FE can select which error it chooses to return. So if a FULLDATA-TLV for a SET/SET-PROP of a structure attempts to write one field which is read only, and attempts to set another field to an invalid value, the FE can return indication of either error. A SET/SET-PROP operation on a variable length component with a length of 0 for the item is not the same as deleting it. If the CE wishes to delete then the DEL operation should be used whether the path refers to an array component or an optional structure component.
The RESULT-TLV is an instance of TLV defined in . The definition is as below:
Defined Result Values Result Value Value Definition E_SUCCESS 0x00 Success E_INVALID_HEADER 0x01 Unspecified error with header. E_LENGTH_MISMATCH 0x02 Header length field does not match actual packet length. E_VERSION_MISMATCH 0x03 Unresolvable mismatch in versions. E_INVALID_DESTINATION_PID 0x04 Destination PID is invalid for the message receiver. E_LFB_UNKNOWN 0x05 LFB Class ID is not known by receiver. E_LFB_NOT_FOUND 0x06 LFB Class ID is known by receiver but not currently in use. E_LFB_INSTANCE_ID_NOT_FOUND 0x07 LFB Class ID is known but the specified instance of that class does not exist. E_INVALID_PATH 0x08 The specified path is impossible. E_COMPONENT_DOES_NOT_EXIST 0x09 The specified path is possible but the component does not exist (e.g., attempt to modify a table row that has not been created). E_EXISTS 0x0A The specified object exists but it cannot exist for the operation to succeed (e.g., attempt to add an existing LFB instance or array subscript). E_NOT_FOUND 0x0B The specified object does not exist but it MUST exist for the operation to succeed (e.g., attempt to delete a non-existing LFB instance or array subscript). E_READ_ONLY 0x0C Attempt to modify a read-only value. E_INVALID_ARRAY_CREATION 0x0D Attempt to create an array with an unallowed subscript. E_VALUE_OUT_OF_RANGE 0x0E Attempt to set a parameter to a value outside of its allowable range. E_CONTENTS_TOO_LONG 0x0D Attempt to write contents larger than the target object space (i.e., exceeding a buffer). E_INVALID_PARAMETERS 0x10 Any other error with data parameters. E_INVALID_MESSAGE_TYPE 0x11 Message type is not acceptable. E_INVALID_FLAGS 0x12 Message flags are not acceptable for the given message type. E_INVALID_TLV 0x13 A TLV is not acceptable for the given message type. E_EVENT_ERROR 0x14 Unspecified error while handling an event. E_NOT_SUPPORTED 0x15 Attempt to perform a valid ForCES operation that is unsupported by the message receiver. E_MEMORY_ERROR 0x16 A memory error occurred while processing a message (no error detected in the message itself) E_INTERNAL_ERROR 0x17 An unspecified error occurred while processing a message (no error detected in the message itself) - 0x18-0xFE Reserved E_UNSPECIFIED_ERROR 0xFF Unspecified error (for when the FE can not decide what went wrong)
A FULLDATA-TLV has "T"= FULLDATA-TLV and a 16-bit Length followed by the data value/contents. Likewise, a SPARSEDATA-TLV has "T" = SPARSEDATA-TLV, a 16-bit Length, followed by the data value/contents. In the case of the SPARSEDATA-TLV, each component in the Value part of the TLV will be further encapsulated in an ILV. Below are the encoding rules for the FULLDATA-TLV and SPARSEDATA-TLVs. is very useful in illustrating these rules: Both ILVs and TLVs MUST be 32 bit aligned. Any padding bits used for the alignment MUST be zero on transmission and MUST be ignored upon reception. FULLDATA-TLVs may be used at a particular path only if every component at that path level is present. In example 1(c) of this concept is illustrated by the presence of all components of the structure S in the FULLDATA-TLV encoding. This requirement holds regardless of whether the fields are fixed or variable length, mandatory or optional. If a FULLDATA-TLV is used, the encoder MUST lay out data for each component in the same order in which the data was defined in the LFB specification. This ensures the decoder is able to retrieve the data. To use the example 1 again in , this implies the encoder/decoder is assumed to have knowledge of how structure S is laid out in the definition. In the case of a SPARSEDATA-TLV, it does not need to be ordered since the "I" in the ILV uniquely identifies the component. Examples 1(a) and 1(b) in illustrate the power of SPARSEDATA-TLV encoding. Inside a FULLDATA-TLV The values for atomic, fixed-length fields are given without any TLV or ILV encapsulation. The values for atomic, variable-length fields are given inside FULLDATA-TLVs. Inside a SPARSEDATA-TLV The values for atomic fields may be given with ILVs (32-bit index, 32-bit length) Any of the FULLDATA-TLVs can contain an ILV but an ILV cannot contain a FULLDATA-TLV. This is because it is hard to disambiguate the ILV since an I is 32 bits and a T is 16 bits. A FULLDATA-TLV can also contain a FULLDATA-TLV for variable sized components. The decoding disambiguation is assumed from rule #3 above.
It is expected that a GET-RESPONSE would satisfy the following: It would have exactly the same path definitions as those sent in the GET. The only difference being a GET-RESPONSE will contain FULLDATA-TLVs. It should be possible to take the same GET-RESPONSE and convert it to a SET successfully by merely changing the T in the operational TLV. There are exceptions to this rule: When a KEY selector is used with a path in a GET operation, that selector is not returned in the GET-RESPONSE; instead the cooked result is returned. Refer to the examples using KEYS to see this. When dumping a whole table in a GET, the GET-RESPONSE that merely edits the T to be SET will end up overwriting the table.
The figure below shows a general layout of the PL PDU. A main header is followed by one or more LFB selections each of which may contain one or more operation.
The figure below shows a more detailed example of the general layout of the operation within a targeted LFB selection. The idea is to show the different nesting levels a path could take to get to the target path.
shows a more concise set of use-cases on how the data encoding is done.
There are two LFBs that are used to control the operation of the ForCES protocol and to interact with FEs and CEs: FE Protocol LFB FE Object LFB Although these LFBs have the same form and interface as other LFBs, they are special in many respects. They have fixed well-known LFB Class and Instance IDs. They are statically defined (no dynamic instantiation allowed) and their status cannot be changed by the protocol: any operation to change the state of such LFBs (for instance, in order to disable the LFB) MUST result in an error. Moreover, these LFBs MUST exist before the first ForCES message can be sent or received. All attributes in these LFBs MUST have pre-defined default values. Finally, these LFBs do not have input or output ports and do not integrate into the intra-FE LFB topology.
The FE Protocol LFB is a logical entity in each FE that is used to control the ForCES protocol. The FE Protocol LFB Class ID is assigned the value 0x2. The FE Protocol LFB Instance ID is assigned the value 0x1. There MUST be one and only one instance of the FE Protocol LFB in an FE. The values of the attributes in the FE Protocol LFB have pre-defined default values that are specified here. Unless explicit changes are made to these values using Config messages from the CE, these default values MUST be used for correct operation of the protocol. The formal definition of the FE Protocol Object LFB can be found in .
FE Protocol capabilities are read-only.
ForCES protocol version(s) supported by the FE
FE Protocol attributes (can be read and set).
Current running version of the ForCES protocol
FE unicast ID
FE multicast ID(s) list - this is a list of multicast IDs that the FE belongs to. These IDs are configured by the CE.
CE heartbeat policy - This policy, along with the parameter 'CE Heartbeat Dead Interval (CE HDI)' as described below defines the operating parameters for the FE to check the CE liveness. The policy values with meanings are listed as below: 0 (default) - This policy specifies that the CE will send a Heartbeat Message to the FE(s) whenever the CE reaches a time interval within which no other PL messages were sent from the CE to the FE(s); refer to and for details. The CE HDI attribute as described below is tied to this policy. 1 - The CE will not generate any HB messages. This actually means CE does not want the FE to check the CE liveness. Others - reserved.
CE Heartbeat Dead Interval (CE HDI) - The time interval the FE uses to check the CE liveness. If FE has not received any messages from CE within this time interval, FE deduces lost connectivity which implies that the CE is dead or the association to the CE is lost. Default value 30 s.
FE heartbeat policy - This policy, along with the parameter 'FE Heartbeat Interval (FE HI)', defines the operating parameters for how the FE should behave so that the CE can deduce its liveness. The policy values and the meanings are: 0 (default) - The FE should not generate any Heartbeat messages. In this scenario, the CE is responsible for checking FE liveness by setting the PL header ACK flag of the message it sends to AlwaysACK. The FE responds to CE whenever CE sends such Heartbeat Request Message. Refer to and for details. 1 - This policy specifies that FE MUST actively send a Heartbeat Message if it reaches the time interval assigned by the FE HI as long as no other messages were sent from FE to CE during that interval as described in . Others - Reserved.
FE Heartbeat Interval (FE HI) - The time interval the FE should use to send HB as long as no other messages were sent from FE to CE during that interval as described in . The default value for an FE HI is 500ms.
Primary CEID - The CEID that the FE is associated with.
Last Primary CEID - The CEID of the last CE that that the FE associated with. This CE ID is reported to the new Primary CEID.
The list of backup CEs an FE can use as backups. Refer to for details.
CE failover policy - This specifies the behavior of the FE when the association with the CE is lost. There is a very tight relation between CE failover policy and , , , and . When an association is lost, depending on configuration, one of the policies listed below is activated. 0 (default) - FE should stop functioning immediately and transition to FE OperDisable. 1 - The FE should continue running and do what it can even without an associated CE. This basically requires that the FE support CE Graceful restart (and defines such support in its capabilities). If the CEFTI expires before the FE re-associates with either the primary () or one of possibly several backup CEs (), the FE will go operationally down. Others - Reserved
CE Failover Timeout Interval (CEFTI) - The time interval associated with the CE failover policy case '0' and '2'. The default value is set to 300 seconds. Note that it is advisable to set the CEFTI value much higher than the CE Heartbeat Dead Interval (CE HDI) since the effect of expiring this parameter is devastating to the operation of the FE.
FE restart policy - This specifies the behavior of the FE during an FE restart. The restart may be from an FE failure or other reasons that have made FE down and then need to restart. The values are defined as below: 0(default)- Restart the FE from scratch. In this case, the FE should start from the pre-association phase. others - Reserved for future use.
The FE Object LFB is a logical entity in each FE and contains attributes relative to the FE itself, and not to the operation of the ForCES protocol. The formal definition of the FE Object LFB can be found in . The model captures the high level properties of the FE that the CE needs to know to begin working with the FE. The class ID for this LFB Class is also assigned in . The singular instance of this class will always exist, and will always have instance ID 0x1 within its class. It is common, although not mandatory, for a CE to fetch much of the component and capability information from this LFB instance when the CE begins controlling the operation of the FE.
Recall: The PL provides a master(CE)-Slave(FE) relationship. The LFBs reside at the FE and are controlled by CE. When messages go from the CE, the LFB Selector (Class and instance) refers to the destination LFB selection which resides in the FE. When messages go from the FE to the CE, the LFB Selector (Class and instance) refers to the source LFB selection which resides in the FE.
The ForCES Association messages are used to establish and teardown associations between FEs and CEs.
This message is sent by the FE to the CE to setup a ForCES association between them. FE to CE The Message Type in the header is set MessageType= 'AssociationSetup'. The ACK flag in the header MUST be ignored, and the association setup message always expects to get a response from the message receiver (CE), whether the setup is successful or not. The correlator field in the header is set, so that FE can correlate the response coming back from the CE correctly. The FE may set the source ID to 0 in the header to request that the CE should assign an FE ID for the FE in the setup response message. The association setup message body optionally consists of zero, one or two LFBselect TLVs, as described in . The Association Setup message only operates on the FE Object and FE Protocol LFBs, therefore, the LFB class ID in the LFBselect TLV only points to these two kinds of LFBs. There is only one FE object LFB and one FE protocol LFB per FE, therefore, the LFB instance IDs for the FE object LFB and the FE protocol LFB are usually assigned to the value 0x1. The OPER-TLV in the LFBselect TLV is defined as a 'REPORT' operation. More than one component may be announced in this message using REPORT operation to let the FE declare its configuration parameters in an unsolicited manner. These may contain components suggesting values such as the FE HB Interval, or the FEID. The OPER-TLV used is defined below.
Only one operation type is defined for the association setup message: Type = "REPORT" - this type of operation is for FE to report something to CE. This is generically a PATH-DATA-TLV format that has been defined in section () in the PATH-DATA BNF definition. The PATH-DATA-TLV for REPORT operation MAY contain FULLDATA-TLV(s) but SHALL NOT contain any RESULT-TLV in the data format. The RESULT-TLV is defined in and the FULLDATA-TLV is defined in . To better illustrate the above PDU format, a tree structure for the format is shown below:
This message is sent by the CE to the FE in response to the Setup message. It indicates to the FE whether the setup is successful or not, i.e., whether an association is established. CE to FE The Message Type in the header is set MessageType= 'AssociationSetupResponse'. The ACK flag in the header MUST be ignored, and the setup response message never expects to get any more responses from the message receiver (FE). The destination ID in the header will be set to the source ID in the corresponding association setup message, unless that source ID was 0. If the corresponding source ID was 0, then the CE will assign an FE ID value and use that value for the destination ID.
The type of the TLV is "ASResult". Length of the TLV including the T and L fields, in octets. This indicates whether the setup msg was successful or whether the FE request was rejected by the CE. the defined values are: 0 = success 1 = FE ID invalid 2 = permission denied To better illustrate the above PDU format, a tree structure for the format is shown below:
This message can be sent by the FE or CE to any ForCES element to end its ForCES association with that element. CE to FE, or FE to CE (or CE to CE) The Message Type in the header is set MessageType= "AssociationTeardown". The ACK flag MUST be ignored. The correlator field in the header MUST be set to zero and MUST be ignored by the receiver.
The type of the TLV is "ASTreason". Length of the TLV including the T and L fields, in octets. This indicates the reason why the association is being terminated. Several reason codes are defined as follows. 0 - normal teardown by administrator 1 - error - loss of heartbeats 2 - error - out of bandwidth 3 - error - out of memory 4 - error - application crash 255 - error - other or unspecified To better illustrate the above PDU format, a tree structure for the format is shown below:
The ForCES Configuration messages are used by CE to configure the FEs in a ForCES NE and report the results back to the CE.
This message is sent by the CE to the FE to configure LFB components in the FE. This message is also used by the CE to subscribe/unsubscribe to LFB events. As usual, a config message is composed of a common header followed by a message body that consists of one or more TLV data format. Detailed description of the message is as below. CE to FE The Message Type in the header is set MessageType= 'Config'. The ACK flag in the header can be set to any value defined in , to indicate whether or not a response from FE is expected by the message.
The operation type for config message. two types of operations for the config message are defined: Type = "SET" - this operation is to set LFB components Type = "SET-PROP" - this operation is to set LFB component properties Type = "DEL" - this operation to delete some LFB components Type = "COMMIT" - this operation is sent to the FE to commit in a 2pc transaction. A COMMIT TLV is an empty TLV i.e it has no "V"alue. In other words, There is a Length of 4 (which is for the header only). Type = "TRCOMP" - this operation is sent to the FE to mark the success from an NE perspective of a 2pc transaction. A TRCOMP TLV is an empty TLV i.e it has no "V"alue. In other words, There is a Length of 4 (which is for the header only). This is generically a PATH-DATA-TLV format that has been defined in section () in the PATH-DATA-TLV BNF definition. The restriction on the use of PATH-DATA-TLV for SET/SET-PROP operation is that it MUST contain either a FULLDATA-TLV or SPARSEDATA-TLV(s), but MUST NOT contain any RESULT-TLV. The restriction on the use of PATH-DATA-TLV for DEL operation is it MAY contain FULLDATA-TLV or SPARSEDATA-TLV(s), but MUST NOT contain any RESULT-TLV. The RESULT-TLV is defined in and FULLDATA-TLV and SPARSEDATA-TLVs is defined in . For Event subscription, the events will be defined by the individual LFBs. To better illustrate the above PDU format, a tree structure for the format is shown below:
This message is sent by the FE to the CE in response to the Config message. It indicates whether the Config was successful or not on the FE and also gives a detailed response regarding the configuration result of each component. FE to CE The Message Type in the header is set MessageType= 'Config Response'. The ACK flag in the header is always ignored, and the Config Response message never expects to get any further response from the message receiver (CE).
The operation type for Config Response message. Two types of operations for the Config Response message are defined: Type = "SET-RESPONSE" - this operation is for the response of SET operation of LFB components Type = "SET-PROP-RESPONSE" - this operation is for the response of SET-PROP operation of LFB component properties Type = "DEL-RESPONSE" - this operation is for the response of the DELETE operation of LFB components Type = "COMMIT-RESPONSE" - this operation is sent to the CE to confirm a commit success in a 2pc transaction. A COMMIT-RESPONSE TLV MUST contain a RESULT-TLV indicating success or failure. This is generically a PATH-DATA-TLV format that has been defined in section () in the PATH-DATA-TLV BNF definition. The restriction on the use of PATH-DATA-TLV for SET-RESPONSE operation is that it MUST contain RESULT-TLV(s). The restriction on the use of PATH-DATA-TLV for DEL-RESPONSE operation is it also MUST contain RESULT-TLV(s). The RESULT-TLV is defined in . To better illustrate the above PDU format, a tree structure for the format is shown below:
The ForCES query messages are used by the CE to query LFBs in the FE for information like LFB components, capabilities, statistics, etc. Query Messages include the Query Message and the Query Response Message.
A Query message is composed of a common header and a message body that consists of one or more TLV data format. Detailed description of the message is as below. from CE to FE The Message Type in the header is set to MessageType= 'Query'. The ACK flag in the header is always ignored, and a full response for a query message is always expected. The Correlator field in the header is set, so that the CE can locate the response back from FE correctly.
The operation type for query. Two operation types are defined: Type = "GET" - this operation is to request to get LFB components. Type = "GET-PROP" - this operation is to request to get LFB components. This is generically a PATH-DATA-TLV format that has been defined in section () in the PATH-DATA-TLV BNF definition. The restriction on the use of PATH-DATA-TLV for GET/GET-PROP operation is it MUST NOT contain any SPARSEDATA-TLV or FULLDATA-TLV and RESULT-TLV in the data format. To better illustrate the above PDU format, a tree structure for the format is shown below:
When receiving a Query message, the receiver should process the message and come up with a query result. The receiver sends the query result back to the message sender by use of the Query Response Message. The query result can be the information being queried if the query operation is successful, or can also be error codes if the query operation fails, indicating the reasons for the failure. A Query Response message is also composed of a common header and a message body consisting of one or more TLVs describing the query result. Detailed description of the message is as below. from FE to CE The Message Type in the header is set to MessageType= 'QueryResponse'. The ACK flag in the header is ignored. As a response itself, the message does not expect a further response.
The operation type for query response. One operation type is defined: Type = "GET-RESPONSE" - this operation is to response to get operation of LFB components. Type = "GET-PROP-RESPONSE" - this operation is to response to GET-PROP operation of LFB components. This is generically a PATH-DATA-TLV format that has been defined in section () in the PATH-DATA-TLV BNF definition. The PATH-DATA-TLV for GET-RESPONSE operation MAY contain SPARSEDATA-TLV, FULLDATA-TLV and/or RESULT-TLV(s) in the data encoding. The RESULT-TLV is defined in and the SPARSEDATA-TLVs and FULLDATA-TLVs are defined in . To better illustrate the above PDU format, a tree structure for the format is shown below:
Event Notification Message is used by FE to asynchronously notify CE of events that happen in the FE. All events that can be generated in an FE are subscribable by the CE. The CE can subscribe to an event via a Config message with SET-PROP operation, where the included path specifies the event, as defined by the LFB Library and described by the FE Model. As usual, an Event Notification Message is composed of a common header and a message body that consists of one or more TLV data format. Detailed description of the message is as below. FE to CE The Message Type in the message header is set to MessageType = 'EventNotification'. The ACK flag in the header MUST be ignored by the CE, and the event notification message does not expect any response from the receiver.
Only one operation type is defined for the event notification message: Type = "REPORT" - this type of operation is for FE to report something to CE. This is generically a PATH-DATA-TLV format that has been defined in section () in the PATH-DATA-TLV BNF definition. The PATH-DATA-TLV for REPORT operation MAY contain FULLDATA-TLV or SPARSEDATA-TLV(s) but MUST NOT contain any RESULT-TLV in the data format. To better illustrate the above PDU format, a tree structure for the format is shown below:
A packet Redirect message is used to transfer data packets between CE and FE. Usually these data packets are control packets but they may be just data-path packets which need further (exception or high-touch) processing. It is also feasible that this message carries no data packets and rather just metadata. The Packet Redirect message data format is formatted as follows: CE to FE or FE to CE The Message Type in the header is set to MessageType= 'PacketRedirect'. This consists of one or more TLVs that contain or describe the packet being redirected. The TLV is specifically a Redirect TLV (with the TLV Type="Redirect"). Detailed data format of a Redirect TLV for packet redirect message is as below:
This is a TLV that specifies meta-data associated with followed redirected data. The TLV is as follows:
This is an Identifier-Length-Value format that is used to describe one meta data. The ILV has the format as:
Where, Meta Data ID is an identifier for the meta data, which is statically assigned by the LFB definition. This is a TLV describing one packet of data to be directed via the redirect operation. The TLV format is as follows:
This field contains the packet that is to be redirected in network byte order. The packet should be 32-bits aligned as is the data for all TLVs. The metadata infers what kind of packet is carried in value field and therefore allows for easy decoding of data encapsulated To better illustrate the above PDU format, a tree structure for the format is shown below:
The Heartbeat (HB) Message is used for one ForCES element (FE or CE) to asynchronously notify one or more other ForCES elements in the same ForCES NE on its liveness. describes the traffic-sensitive approach used. A Heartbeat Message is sent by a ForCES element periodically. The parameterization and policy definition for heartbeats for an FE is managed as components of the FE Protocol Object LFB, and can be set by CE via a Config message. The Heartbeat message is a little different from other protocol messages in that it is only composed of a common header, with the message body left empty. A detailed description of the message is as below. FE to CE or CE to FE The Message Type in the message header is set to MessageType = 'Heartbeat'. describes the HB mechanisms used. The ACK flag in the header MUST be set to either 'NoACK' or 'AlwaysACK' when the HB is sent. When set to 'NoACK', the HB is not soliciting for a response. When set to 'AlwaysACK', the HB Message sender is always expecting a response from its receiver. According the HB policies defined in , only the CE can send such an HB message to query FE liveness. For simplicity and because of the minimal nature of the HB message, the response to a HB message is another HB message, i.e., no specific HB response message is defined. Whenever an FE receives a HB message marked with 'AlwaysACK' from the CE, the FE MUST send a HB message back immediately. The HB message sent by the FE in response to the 'AlwasyACK' MUST modify the source and destination IDs so that the ID of the FE is the source ID and the CE ID of the sender is the destination ID, and MUST change the ACK information to 'NoACK'. A CE MUST NOT respond to an HB message with 'AlwasyACK' set. When set to anything else other than 'NoACK' or 'AlwaysACK', the HB Message is treated as if it was a 'NoACK'. The correlator field in the HB message header SHOULD be set accordingly when a response is expected so that a receiver can correlate the response correctly. The correlator field MAY be ignored if no response is expected. The message body is empty for the Heartbeat Message.
The ForCES protocol provides mechanisms for CE redundancy and failover, in order to support High Availability as defined in . FE redundancy and FE to FE interaction is currently out of scope of this document. There can be multiple redundant CEs and FEs in a ForCES NE. However, at any one time only one primary CE can control the FEs though there can be multiple secondary CEs. The FE and the CE PL are aware of the primary and secondary CEs. This information (primary, secondary CEs) is configured in the FE and in the CE PLs during pre-association by the FEM and the CEM respectively. Only the primary CE sends control messages to the FEs.
High Availability parameterization in an FE is driven by configuring the FE Protocol Object LFB (refer to and ). The FE Heartbeat Interval, CE Heartbeat Dead Interval, and CE Heartbeat policy help in detecting connectivity problems between an FE and CE. The CE Failover policy defines the reaction on a detected failure. extends the state machine illustrated in to allow for new states that facilitate connection recovery.
-->-->| in +<----+ | | progress) | | | CE Issues +--------+--------+ | | Association | | CFTI | Setup V | timer | ___________________+ | expires | | | | V ^ +-+-----------+ +-------+-----+ | | | Not | | | (CE issues Teardown || | Associated | | | Lost association) && | | | Associated | CE Failover Policy = 1 | (May | | | | Continue | | |---------->------->------>| Forwarding)| | | | | +-------------+ +-------------+ ^ V | | | CE Issues | | Association | | Setup | +_________________________________________+ ]]>
describes transitions between the Pre-association, associated and not associated states. When communication fails between the FE and CE (which can be caused by either the CE or link failure but not FE related), either the TML on the FE will trigger the FE PL regarding this failure or it will be detected using the HB messages between FEs and CEs. The communication failure, regardless of how it is detected, MUST be considered as a loss of association between the CE and corresponding FE. If the FE's FEPO CE Failover Policy is configured to mode 0 (the default), it will immediately transition to the pre-association phase. This means that if association is again established, all FE state will need to be re-established. If the FE's FEPO CE Failover Policy is configured to mode 1, it indicates that the FE is capable of HA restart recovery. In such a case, the FE transitions to the not associated state and the CEFTI timer is started. The FE MAY continue to forward packets during this state. It MAY also recycle through any configured secondary CEs in a round-robin fashion. It first adds its primary CE to the tail of backup CEs and sets its primary CE to be the first secondary. It then attempts to associate with the CE designated as the new primary CE. If it fails to re-associate with any CE and the CEFTI expires, the FE then transitions to the pre-association state. If the FE, while in the not associated state, manages to reconnect to a new primary CE before CEFTI expires it transitions to the Associated state. Once re-associated, the FE tries to recover any state that may have been lost during the not associated state. How the FE achieves re-synchronizes it state is out of scope for this document. below illustrates the Forces message sequences that the FE uses to recover the connection.
| | | | | | All msgs | | 2 |<--------------------->| | | | | | | | | FAILURE | | | | Asso Estb,Caps exchange | 3 |<------------------------------------------>| | | | Event Report (pri CE down) | 4 |------------------------------------------->| | | | All Msgs | 5 |<------------------------------------------>| ]]>
A CE-to-CE synchronization protocol would be needed to support fast failover as well as to address some of the corner cases, however this will not be defined by the ForCES protocol as it is out of scope for this specification. An explicit message (a Config message setting Primary CE component in ForCES Protocol object) from the primary CE, can also be used to change the Primary CE for an FE during normal protocol operation. Also note that the FEs in a ForCES NE could also use a multicast CE ID, i.e., they could be associated with a group of CEs (this assumes the use of a CE-CE synchronization protocol, which is out of scope for this specification). In this case, the loss of association would mean that communication with the entire multicast group of CEs has been lost. The mechanisms described above will apply for this case as well during the loss of association. If, however, the secondary CE was also using the multicast CE ID that was lost, then the FE will need to form a new association using a different CE ID. If the capability exists, the FE MAY first attempt to form a new association with original primary CE using a different non multicast CE ID.
TML Level: The TML controls logical connection availability and failover. The TML also controls peer HA management. At this level, control of all lower layers, for example transport level (such as IP addresses, MAC addresses etc) and associated links going down are the role of the TML. PL Level: All other functionality, including configuring the HA behavior during setup, the CE IDs used to identify primary and secondary CEs, protocol messages used to report CE failure (Event Report), Heartbeat messages used to detect association failure, messages to change the primary CE (Config), and other HA related operations described before, are the PL responsibility. To put the two together, if a path to a primary CE is down, the TML would take care of failing over to a backup path, if one is available. If the CE is totally unreachable then the PL would be informed and it would take the appropriate actions described before.
ForCES Framework document , section 8 goes into extensive detail on a variety of security threats, the possible effects of those threats on the protocol and responses to those threats. This document does not repeat that discussion, the reader is referred to the ForCES Framework document for those details and how the ForCES architecture addresses them. ForCES PL uses security services provided by the ForCES TML. The TML provides security services such as endpoint authentication service, message authentication service and confidentiality service. Endpoint authentication service is invoked at the time of the pre-association connection establishment phase and message authentication is performed whenever the FE or CE receives a packet from its peer. The following are the general security mechanisms that need to be in place for ForCES PL. Security mechanisms are session controlled - that is, once the security is turned on depending upon the chosen security level (No Security, Authentication, Confidentiality), it will be in effect for the entire duration of the session. An operator should configure the same security policies for both primary and backup FEs and CEs (if available). This will ensure uniform operations and avoid unnecessary complexity in policy configuration.
When "No security" is chosen for ForCES protocol communication, both endpoint authentication and message authentication service needs to be performed by ForCES PL. Both these mechanism are weak and do not involve cryptographic operation. An operator can choose "No Security" level when the ForCES protocol endpoints are within a single box, for example. In order to have interoperable and uniform implementation across various security levels, each CE and FE endpoint MUST implement this level. What is described below (in and ) are error checks and not security procedures. The reader is referred to section . for security procedures.
Each CE and FE PL maintains a list of associations as part its of configuration. This is done via the CEM and FEM interfaces. An FE MUST connect to only those CEs that are configured via the FEM; similarly, a CE should accept the connection and establish associations for the FEs which are configured via the CEM. The CE should validate the FE identifier before accepting the connections during the pre-association phase.
When a CE or FE initiates a message, the receiving endpoint MUST validate the initiator of the message by checking the common header CE or FE identifiers. This will ensure proper protocol functioning. This extra processing step is recommended even when the underlying TML layer security services exist.
This section is applicable if an operator wishes to use the TML security services. A ForCES TML MUST support one or more security services such as endpoint authentication service, message authentication service, and confidentiality service, as part of TML security layer functions. It is the responsibility of the operator to select an appropriate security service and configure security policies accordingly. The details of such configuration are outside the scope of the ForCES PL and are dependent on the type of transport protocol and the nature of the connection. All these configurations should be done prior to starting the CE and FE. When certificates-based authentication is being used at the TML, the certificate can use a ForCES-specific naming structure as certificate names and, accordingly, the security policies can be configured at the CE and FE. The reader is asked to refer to specific TML documents for details on the security requirements specific to that TML
When TML security services are enabled, the ForCES TML performs endpoint authentication. Security association is established between CE and FE and is transparent to the ForCES PL.
This is a TML specific operation and is transparent to the ForCES PL. For details, refer to .
This is a TML specific operation and is transparent to the ForCES PL. For details, refer to .
The authors of this draft would like to acknowledge and thank the ForCES Working Group and especially the following: Furquan Ansari, Alex Audu, Steven Blake, Shuchi Chawla, Alan DeKok, Ellen M. Deleganes, Xiaoyi Guo, Yunfei Guo, Evangelos Haleplidis, Joel M. Halpern (who should probably be listed among the authors), Zsolt Haraszti, Fenggen Jia, John C. Lin, Alistair Munro, Jeff Pickering, T. Sridhlar, Guangming Wang, Chaoping Wu, and Lily L. Yang, for their contributions. We would also like to thank David Putzolu, and Patrick Droz for their comments and suggestions on the protocol and for their infinite patience. We would also like to thank Sue Hares and Alia Atlas for extensive reviews of the document. Alia Atlas has done a wonderful job of shaping the draft to make it more readable by providing the IESG feedback. The editors have used the xml2rfc tools in creating this document and are very grateful for the existence and quality of these tools. The editor is also grateful to Elwyn Davies for his help in correcting the XML source of this document.
&RFC2119; &RFC2914; &RFC5390; &RFC5226; ForCES Forwarding Element Model SCTP based TML (Transport Mapping Layer) for ForCES protocol &RFC2629; &RFC3654; &RFC3746; Principles of Transaction-Orientated Database Recovery Notes on database operating systems. In Operating Systems: An Advanced Course. Lecture Notes in Computer Science, Vol. 60, pp. 394-481, Springer-Verlag
Following the policies outlined in "Guidelines for Writing an IANA Considerations Section in RFCs" (RFC 5226 ), the following name spaces are defined in ForCES. Message Type Name Space Operation Type Name Space Header Flags TLV Type TLV Result Values LFB Class ID Result: Association Setup Response Reason: Association Teardown Message
The Message Type is an 8 bit value. The following is the guideline for defining the Message Type namespace Message Types in this range are part of the base ForCES Protocol. Message Types in this range are allocated through an IETF consensus action. Values assigned by this specification:
Message Types in this range are Specification Required Message Types using this range MUST be documented in an RFC or other permanent and readily available reference. Message Types in this range are reserved for vendor private extensions and are the responsibility of individual vendors. IANA management of this range of the Message Type Name Space is unnecessary.
The Operation Selection (OPER-TLV) name space is 16 bits long. The following is the guideline for managing the OPER-TLV Name Space. OPER-TLV Types in this range are allocated through an IETF consensus process. . Values assigned by this specification:
OPER-TLV Types using this range MUST be documented in an RFC or other permanent and readily available reference. . OPER-TLV Types in this range are reserved for vendor private extensions and are the responsibility of individual vendors. IANA management of this range of the OPER-TLV Type Name Space is unnecessary.
The Header flag field is 32 bits long. Header flags are part of the ForCES base protocol. Header flags are allocated through an IETF consensus action .
The TLV Type name space is 16 bits long. The following is the guideline for managing the TLV Type Name Space. TLV Types in this range are allocated through an IETF consensus process. . Values assigned by this specification:
TLV Types using this range MUST be documented in an RFC or other permanent and readily available reference . TLV Types in this range are reserved for vendor private extensions and are the responsibility of individual vendors. IANA management of this range of the TLV Type Name Space is unnecessary.
The RESULT-TLV RTesult Value is an 8 bit value.
All values not assigned in this specification are designated as Assignment by Expert review.
The Association Setup Response name space is 32 bits long. The following is the guideline for managing the Association Setup Response Name Space. Association Setup Responses in this range are allocated through an IETF consensus process . Values assigned by this specification:
Association Setup Responses in this range are Specification Required Values using this range MUST be documented in an RFC or other permanent and readily available reference . Association Setup Responses in this range are reserved for vendor private extensions and are the responsibility of individual vendors. IANA management of this range of the Association Setup Responses Name Space is unnecessary.
The Association Teardown Message name space is 32 bits long. The following is the guideline for managing the Association Teardown Message Name Space. Association Teardown Messages in this range are allocated through an IETF consensus process . Values assigned by this specification:
Association Teardown Messages in this range are Specification Required Association Teardown Messages using this range MUST be documented in an RFC or other permanent and readily available references. . Association Teardown Messages in this range are reserved for vendor private extensions and are the responsibility of individual vendors. IANA management of this range of the Association Teardown Message Name Space is unnecessary.
The schema described below conforms to the LFB schema described in ForCES Model draft. describes the details of the different components defined in this definition.
CEHBPolicyValues The possible values of CE heartbeat policy uchar CEHBPolicy0 The CE heartbeat policy 0 CEHBPolicy1 The CE heartbeat policy 1 FEHBPolicyValues The possible values of FE heartbeat policy uchar FEHBPolicy0 The FE heartbeat policy 0 FEHBPolicy1 The FE heartbeat policy 1 FERestartPolicyValues The possible values of FE restart policy uchar FERestartPolicy0 The FE restart policy 0 CEFailoverPolicyValues The possible values of CE failover policy uchar CEFailoverPolicy0 The CE failover policy 0 CEFailoverPolicy1 The CE failover policy 1 FEHACapab The supported HA features uchar GracefullRestart The FE supports Graceful Restart HA The FE supports HA FEPO The FE Protocol Object 1.0 CurrentRunningVersion Currently running ForCES version u8 FEID Unicast FEID uint32 MulticastFEIDs the table of all multicast IDs uint32 CEHBPolicy The CE Heartbeat Policy CEHBPolicyValues CEHDI The CE Heartbeat Dead Interval in millisecs uint32 FEHBPolicy The FE Heartbeat Policy FEHBPolicyValues FEHI The FE Heartbeat Interval in millisecs uint32 CEID The Primary CE this FE is associated with uint32 BackupCEs The table of all backup CEs other than the primary uint32 CEFailoverPolicy The CE Failover Policy CEFailoverPolicyValues CEFTI The CE Failover Timeout Interval in millisecs uint32 FERestartPolicy The FE Restart Policy FERestartPolicyValues LastCEID The Primary CE this FE was last associated with uint32 SupportableVersions the table of ForCES versions that FE supports u8 HACapabilities the table of HA capabilities the FE supports FEHACapab PrimaryCEDown The pimary CE has changed LastCEID LastCEID ]]>
> Supportable Versions enumerates all ForCES versions that an FE supports. FEHACapab enumerates the HA capabilities of the FE. If the FE is not capable of Graceful restarts or HA, then it will not be able to participate in HA as described in
> All Components are explained in .
In this section a few examples of data encoding are discussed. these example, however, do not show any padding.
Structure with three fixed-lengthof, mandatory fields.
(a) Describing all fields using SPARSEDATA-TLV
(b) Describing a subset of fields
Note: Even though there are non-optional components in structure S, since one can uniquely identify components, one can selectively send component of structure S (eg in the case of an update from CE to FE). (c) Describing all fields using a FULLDATA-TLV
Structure with three fixed-lengthof fields, one mandatory, two optional.
This example is identical to Example 1, as illustrated below. (a) Describing all fields using SPARSEDATA-TLV
(b) Describing a subset of fields using SPARSEDATA-TLV
(c) Describing all fields using a FULLDATA-TLV
Note: FULLDATA-TLV _cannot_ be used unless all fields are being described.
Structure with a mix of fixed-lengthof and variable-lengthof fields, some mandatory, some optional. Note in this case, b is variable sized
(a) Describing all fields using SPARSEDATA-TLV
(b) Describing a subset of fields using SPARSEDATA-TLV
(c) Describing all fields using FULLDATA-TLV
Note: The variable-length field requires the addition of a FULLDATA-TLV within the outer FULLDATA-TLV as in the case of component b above.
Structure containing an array of another structure type.
(a) Encoding using SPARSEDATA-TLV, with two instances of z[], also described with SPARSEDATA-TLV, assuming only the 10th and 15th subscript of z[] are encoded.
Note the holes in the components of z (10 followed by 15). Also note the gap in index 15 with only components a and c appearing but not b.
Assume LFB with following components for the following use cases.
All examples will use valueof(x) to indicate the value of the referenced component x. In the case where F_SEL** are missing (bits equal to 00) then the flags will not show any selection. All the examples only show use of FULLDATA-TLV for data encoding; although SPARSEDATA-TLV would make more sense in certain occasions, the emphasis is on showing the message layout. Refer to for examples that show usage of both FULLDATA-TLV and SPARSEDATA-TLV. To get foo1
To set foo2 to 10
To dump table2
One should be able to take a GET-RESPONSE-TLV and convert it to a SET-TLV. If the result in the above example is sent back in a SET-TLV, (instead of a GET-RESPONSE_TLV) then the entire contents of the table will be replaced at that point. Multiple operations Example. To create entry 0-5 of table2 (Error conditions are ignored)
Block operations (with holes) example. Replace entry 0,2 of table2
Getting rows example. Get first entry of table2.
Get entry 0-5 of table2.
Create a row in table2, index 5.
Dump contents of table1.
Using Keys. Get row entry from table4 where j1=100. Recall, j1 is a defined key for this table and its KeyID is 1.
Delete row with KEY match (j1=100, j2=200) in table2. Note that the j1,j2 pair are a defined key for the table2.
Dump contents of table3. It should be noted that this table has a column with component name that is variable sized. The purpose of this use case is to show how such an component is to be encoded.
Multiple atomic operations. This emulates adding a new nexthop entry and then atomically updating the L3 entries pointing to an old NH to point to a new one. The assumption is both tables are in the same LFB Observe the two operations on the LFB instance, both are SET operations.
Selective setting. On table4 -- for indices 1, 3, 5, 7, and 9. Replace j1 to 100, j2 to 200, j3 to 300. Leave j4 as is.
Manipulation of table of table examples. Get x1 from table10 row with index 4, inside table5 entry 10
From table5's row 10 table10, get X2s based on on the value of x1 equaling 10 (recall x1 is KeyID 1)
Further example of manipulating a table of tables
Get a whole LFB (all its components, etc.). at startup a CE might well want the entire FE OBJECT LFB. So, in a request targeted at class 1, instance 1, one might find: