This Internet-Draft is submitted to IETF in full conformance with the provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet-Drafts.
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as “work in progress.”
The list of current Internet-Drafts can be accessed at http://www.ietf.org/ietf/1id-abstracts.txt.
The list of Internet-Draft Shadow Directories can be accessed at http://www.ietf.org/shadow.html.
This Internet-Draft will expire on March 22, 2010.
Copyright (c) 2009 IETF Trust and the persons identified as the document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents in effect on the date of publication of this document (http://trustee.ietf.org/license-info). Please review these documents carefully, as they describe your rights and restrictions with respect to this document.
This document provides a framework for the development of IP fast-reroute mechanisms which provide protection against link or router failure by invoking locally determined repair paths. Unlike MPLS fast-reroute, the mechanisms are applicable to a network employing conventional IP routing and forwarding.
3. Problem Analysis
4. Mechanisms for IP Fast-reroute
4.1. Mechanisms for fast failure detection
4.2. Mechanisms for repair paths
4.2.1. Scope of repair paths
4.2.2. Analysis of repair coverage
4.2.3. Link or node repair
4.2.4. Maintenance of Repair paths
4.2.5. Multiple failures and Shared Risk Link Groups
4.3. Local Area Networks
4.4. Mechanisms for micro-loop prevention
5. Management Considerations
6. Scope and applicability
7. IANA Considerations
8. Security Considerations
10. Informative References
§ Authors' Addresses
This section defines words and acronyms used in this draft and other drafts discussing IP fast-reroute.
- Used to denote the destination router under discussion.
- The distance of the shortest path from A to B.
- Downstream Path
- This is a subset of the loop-free alternates where the neighbor N meets the following condition:-
Distance_opt(N, D) < Distance_opt(S,D)
- Used to denote the router which is the primary next-hop neighbor to get from S to the destination D. Where there is an ECMP set for the shortest path from S to D, these are referred to as E_1, E_2, etc.
- Equal cost multi-path: Where, for a particular destination D, multiple primary next-hops are used to forward traffic because there exist multiple shortest paths from S via different output layer-3 interfaces.
- Forwarding Information Base. The database used by the packet forwarder to determine what actions to perform on a packet.
- IP fast-reroute.
- A link connecting router A to router B.
- Loop Free Alternate. A neighbor N, that is not a primary next-hop neighbor E, whose shortest path to the destination D does not go back through the router S. The neighbor N must meet the following condition:-
Distance_opt(N, D) < Distance_opt(N, S) + Distance_opt(S, D)
- Loop Free Neighbor
- A neighbor N_i, which is not the particular primary neighbor E_k under discussion, and whose shortest path to D does not traverse S. For example, if there are two primary neighbors E_1 and E_2, E_1 is a loop-free neighbor with regard to E_2 and vice versa.
- Loop Free Link Protecting Alternate
- A path via a Loop-Free Neighbor N_i that reaches destination D without going through the particular link of S that is being protected. In some cases the path to D may go through the primary neighbor E.
- Loop Free Node-protecting Alternate
- A path via a Loop-Free Neighbor N_i that reaches destination D without going through the particular primary neighbor (E) of S which is being protected.
- The ith neighbor of S.
- Primary Neighbor
- A neighbor N_i of S which is one of the next hops for destination D in S's FIB prior to any failure.
- The jth neighbor of N_i.
- Routing Transition
- The process whereby routers converge on a new topology. In conventional networks this process frequently causes some disruption to packet delivery.
- Reverse Path Forwarding. I.e. checking that a packet is received over the interface which would be used to send packets addressed to the source address of the packet.
- Used to denote a router that is the source of a repair that is computed in anticipation of the failure of a neighboring router denoted as E, or of the link between S and E. It is the viewpoint from which IP fast-reroute is described.
- Shortest Path First, e.g. Dijkstra's algorithm.
- Shortest path tree
- Upstream Forwarding Loop
- A forwarding loop that involves a set of routers, none of which are directly connected to the link that has caused the topology change that triggered a new SPF in any of the routers.
When a link or node failure occurs in a routed network, there is inevitably a period of disruption to the delivery of traffic until the network re-converges on the new topology. Packets for destinations which were previously reached by traversing the failed component may be dropped or may suffer looping. Traditionally such disruptions have lasted for periods of at least several seconds, and most applications have been constructed to tolerate such a quality of service.
Recent advances in routers have reduced this interval to under a second for carefully configured networks using link state IGPs. However, new Internet services are emerging which may be sensitive to periods of traffic loss which are orders of magnitude shorter than this.
Addressing these issues is difficult because the distributed nature of the network imposes an intrinsic limit on the minimum convergence time which can be achieved.
However, there is an alternative approach, which is to compute backup routes that allow the failure to be repaired locally by the router(s) detecting the failure without the immediate need to inform other routers of the failure. In this case, the disruption time can be limited to the small time taken to detect the adjacent failure and invoke the backup routes. This is analogous to the technique employed by MPLS fast-reroute [RFC4090] (Pan, P., Swallow, G., and A. Atlas, “Fast Reroute Extensions to RSVP-TE for LSP Tunnels,” May 2005.), but the mechanisms employed for the backup routes in pure IP networks are necessarily very different.
This document provides a framework for the development of this approach.
The duration of the packet delivery disruption caused by a conventional routing transition is determined by a number of factors:
The disruption will last until the routers adjacent to the failure have completed steps 1 and 2, and then all the routers in the network whose paths are affected by the failure have completed the remaining steps.
The initial packet loss is caused by the router(s) adjacent to the failure continuing to attempt to transmit packets across the failure until it is detected. This loss is unavoidable, but the detection time can be reduced to a few tens of milliseconds as described in Section 4.1 (Mechanisms for fast failure detection ).
In some topologies subsequent packet loss may be caused by the "micro-loops" which may form as a result of temporary inconsistencies between routers' forwarding tables[I‑D.ietf‑rtgwg‑lf‑conv‑frmwk] (Shand, M. and S. Bryant, “A Framework for Loop-free Convergence,” October 2009.). These inconsistencies are caused by steps 3, 4 and 5 above and in many routers it is step 5 which is both the largest factor and which has the greatest variance between routers. The large variance arises from implementation differences and from the differing impact that a failure has on each individual router. For example, the number of prefixes affected by the failure may vary dramatically from one router to another.
In order to achieve packet disruption times which are commensurate with the failure detection times two mechanisms may be required:-
Performing the first task without the second may result in the repair path being starved of traffic and hence being redundant. Performing the second without the first will result in traffic being discarded by the router(s) adjacent to the failure.
Repair paths may always be used in isolation where the failure is short-lived. In this case, the repair paths can be kept in place until the failure is repaired in which case there is no need to advertise the failure to other routers.
Similarly, micro-loop avoidance may be used in isolation to prevent loops arising from pre-planned management action. In which case the link or node being shut down can remain in service for a short time after its removal has been announced into the network, and hence it can function as its own "repair path".
Note that micro-loops may also occur when a link or node is restored to service and thus a micro-loop avoidance mechanism may be required for both link up and link down cases.
The set of mechanisms required for an effective solution to the problem can be broken down into the sub-problems described in this section.
It is critical that the failure detection time is minimized. A number of well documented approaches are possible, such as:
Once a failure has been detected by one of the above mechanisms, traffic which previously traversed the failure is transmitted over one or more repair paths. The design of the repair paths should be such that they can be pre-calculated in anticipation of each local failure and made available for invocation with minimal delay. There are three basic categories of repair paths:
ECMP and loop free alternate paths (as described in [RFC5286] (Atlas, A. and A. Zinin, “Basic Specification for IP Fast Reroute: Loop-Free Alternates,” September 2008.)) offer the simplest repair paths and would normally be used when they are available. It is anticipated that around 80% of failures (see Section 4.2.2 (Analysis of repair coverage )) can be repaired using these basic methods alone.
Multi-hop repair paths are more complex, both in the computations required to determine their existence, and in the mechanisms required to invoke them. They can be further classified as:
In many cases a repair path which reaches two hops away from the router detecting the failure will suffice, and it is anticipated that around 98% of failures (see Section 4.2.2 (Analysis of repair coverage )) can be repaired by this method. However, to provide complete repair coverage some use of longer multi-hop repair paths is generally necessary.
A particular repair path may be valid for all destinations which require repair or may only be valid for a subset of destinations. If a repair path is valid for a node immediately downstream of the failure, then it will be valid for all destinations previously reachable by traversing the failure. However, in cases where such a repair path is difficult to achieve because it requires a high order multi-hop repair path, it may still be possible to identify lower order repair paths (possibly even loop free alternate paths) which allow the majority of destinations to be repaired. When IPFRR is unable to provide complete repair, it is desirable that the extent of the repair coverage can be determined and reported via network management.
There is a trade-off to be achieved between minimizing the number of repair paths to be computed, and minimizing the overheads incurred in using higher order multi-hop repair paths for destinations for which they are not strictly necessary. However, the computational cost of determining repair paths on an individual destination basis can be very high.
It will frequently be the case that the majority of destinations may be repaired using only the "basic" repair mechanism, leaving a smaller subset of the destinations to be repaired using one of the more complex multi-hop methods. Such a hybrid approach may go some way to resolving the conflict between completeness and complexity.
The use of repair paths may result in excessive traffic passing over a link, resulting in congestion discard. This reduces the effectiveness of IPFRR. Mechanisms to influence the distribution of repaired traffic to minimize this effect are therefore desirable.
The repair coverage obtained is dependent on the repair strategy and highly dependent on the detailed topology and metrics. Estimates of the repair coverage quoted in this document are for illustrative purposes only and may not be always be achievable.
In some cases the repair strategy will permit the repair of all single link or node failures in the network for all possible destinations. This can be defined as 100% coverage. However, where the coverage is less than 100% it is important for the purposes of comparisons between different proposed repair strategies to define what is meant by such a percentage. There are four possibilities:
A repair path may be computed to protect against failure of an adjacent link, or failure of an adjacent node. In general, link protection is simpler to achieve. A repair which protects against node failure will also protect against link failure for all destinations except those for which the adjacent node is a single point of failure.
In some cases it may be necessary to distinguish between a link or node failure in order that the optimal repair strategy is invoked. Methods for link/node failure determination may be based on techniques such as BFD[I‑D.ietf‑bfd‑base] (Katz, D. and D. Ward, “Bidirectional Forwarding Detection,” January 2010.). This determination may be made prior to invoking any repairs, but this will increase the period of packet loss following a failure unless the determination can be performed as part of the failure detection mechanism itself. Alternatively, a subsequent determination can be used to optimise an already invoked default strategy.
In order to meet the response time goals, it is expected (though not required) that repair paths, and their associated FIB entries, will be pre-computed and installed ready for invocation when a failure is detected. Following invocation the repair paths remain in effect until they are no longer required. This will normally be when the routing protocol has re-converged on the new topology taking into account the failure, and traffic will no longer be using the repair paths.
The repair paths have the property that they are unaffected by any topology changes resulting from the failure which caused their instantiation. Therefore there is no need to re-compute them during the convergence period. They may be affected by an unrelated simultaneous topology change, but such events are out of scope of this work (see Section 4.2.5 (Multiple failures and Shared Risk Link Groups )).
Once the routing protocol has re-converged it is necessary for all repair paths to take account of the new topology. Various optimizations may permit the efficient identification of repair paths which are unaffected by the change, and hence do not require full re-computation. Since the new repair paths will not be required until the next failure occurs, the re-computation may be performed as a background task and be subject to a hold-down, but excessive delay in completing this operation will increase the risk of a new failure occurring before the repair paths are in place.
Complete protection against multiple unrelated failures is out of scope of this work. However, it is important that the occurrence of a second failure while one failure is undergoing repair should not result in a level of service which is significantly worse than that which would have been achieved in the absence of any repair strategy.
Shared Risk Link Groups (SRLGs) are an example of multiple related failures, and the more complex aspects of their protection is a matter for further study.
One specific example of an SRLG which is clearly within the scope of this work is a node failure. This causes the simultaneous failure of multiple links, but their closely defined topological relationship makes the problem more tractable.
Protection against partial or complete failure of LANs is more complex than the point to point case. In general there is a trade-off between the simplicity of the repair and the ability to provide complete and optimal repair coverage.
Ensuring the absence of micro-loops is important not only because they can cause packet loss in traffic which is affected by the failure, but because by saturating a link with looping packets they can also cause congestion loss of traffic flowing over that link which would otherwise be unaffected by the failure.
A number of solutions to the problem of micro-loop formation have been proposed and are summarized in [I‑D.ietf‑rtgwg‑lf‑conv‑frmwk] (Shand, M. and S. Bryant, “A Framework for Loop-free Convergence,” October 2009.). The following factors are significant in their classification:
While many of the management requirements will be specific to particular IPFRR solutions, the following general aspects need to be addressed:
- Enabling/disabling IPFRR support.
- Enabling/disabling protection on a per link/node basis.
- Expressing preferences regarding the links/nodes used for repair paths.
- Configuration of failure detection mechanisms.
- Configuration of loop avoidance strategies
- Notification of links/nodes/destinations which cannot be protected.
- Notification of pre-computed repair paths, and anticipated traffic patterns.
- Counts of failure detections, protection invocations and packets forwarded over repair paths.
- Testing repairs.
The initial scope of this work is in the context of link state IGPs. Link state protocols provide ubiquitous topology information, which facilitates the computation of repairs paths.
Provision of similar facilities in non-link state IGPs and BGP is a matter for further study, but the correct operation of the repair mechanisms for traffic with a destination outside the IGP domain is an important consideration for solutions based on this framework
There are no IANA considerations that arise from this framework document.
This framework document does not itself introduce any security issues, but attention must be paid to the security implications of any proposed solutions to the problem.
Where the chosen solution uses tunnels it is necessary to ensure that the tunnel is not used as an attack vector. One method of addressing this is to use a set of tunnel endpoint addresses that are excluded from use by user traffic.
There is a compatibility issue between IPFRR and reverse path forwarding (RPF) checking. Many of the solutions described in this document result in traffic arriving from a direction inconsistent with a standard RPF check. When a network relies on RPF checking for security purposes, an alternative security mechanism will need to be deployed in order to permit IPFRR to used.
Because the repair path will often be of a different length to the pre-failure path, security mechanisms which rely on specific TTL values will be adversely affected.
The authors would like to acknowledge contributions made by Alia Atlas, Clarence Filsfils, Pierre Francois, Joel Halpern, Stefano Previdi and Alex Zinin.
|[FIFR]||Nelakuditi, S., Lee, S., Lu, Y., Zhang, Z., and C. Chuah, “Fast local rerouting for handling transient link failures.",” Tech. Rep. TR-2004-004, 2004.|
|[I-D.atlas-ip-local-protect-uturn]||Atlas, A., “U-turn Alternates for IP/LDP Fast-Reroute,” draft-atlas-ip-local-protect-uturn-03 (work in progress), March 2006 (TXT).|
|[I-D.bryant-ipfrr-tunnels]||Bryant, S., Filsfils, C., Previdi, S., and M. Shand, “IP Fast Reroute using tunnels,” draft-bryant-ipfrr-tunnels-03 (work in progress), November 2007 (TXT).|
|[I-D.ietf-bfd-base]||Katz, D. and D. Ward, “Bidirectional Forwarding Detection,” draft-ietf-bfd-base-11 (work in progress), January 2010 (TXT).|
|[I-D.ietf-rtgwg-ipfrr-notvia-addresses]||Shand, M., Bryant, S., and S. Previdi, “IP Fast Reroute Using Not-via Addresses,” draft-ietf-rtgwg-ipfrr-notvia-addresses-05 (work in progress), March 2010 (TXT).|
|[I-D.ietf-rtgwg-lf-conv-frmwk]||Shand, M. and S. Bryant, “A Framework for Loop-free Convergence,” draft-ietf-rtgwg-lf-conv-frmwk-07 (work in progress), October 2009 (TXT).|
|[I-D.tian-frr-alt-shortest-path]||Tian, A., “Fast Reroute using Alternative Shortest Paths,” draft-tian-frr-alt-shortest-path-01 (work in progress), July 2004 (TXT).|
|[RFC4090]||Pan, P., Swallow, G., and A. Atlas, “Fast Reroute Extensions to RSVP-TE for LSP Tunnels,” RFC 4090, May 2005 (TXT).|
|[RFC5286]||Atlas, A. and A. Zinin, “Basic Specification for IP Fast Reroute: Loop-Free Alternates,” RFC 5286, September 2008 (TXT).|
|[SIMULA]||Lysne, O., Kvalbein, A., Cicic, T., Gjessing, S., and A. Hansen, “Fast IP Network Recovery using Multiple Routing Configurations.",” Infocom 10.1109/INFOCOM.2006.227, 2006.|
|250, Longwater Avenue.|
|Reading, Berks RG2 6GB|
|250, Longwater Avenue.|
|Reading, Berks RG2 6GB|