05 - Ethernet CFM Across Domains - A List of Open Issues with Multi-Domain Fault Management

Freek Dijkstra, Ronald van der Pol, Sander Boele (SARA)

Before Ethernet 802.1ag/Y.1731 Connectivity Fault Management (CFM) can be deployed for Ethernet circuits spanning multiple research networks, the following seven choices need to be made.
1. Continuity Check or Link Trace
Both techniques can be used to monitor inter-domain Ethernet circuits. Continuity checks are more suited to testing the link status, while link traces are superior in localisation of a fault. We recommend to deploy both technologies.
2. Who deploys it
Ping, traceroute and iperf are currently deployed at end-sites (campus networks), while link status monitoring tools are deployed by network providers (NRENs). Continuity check measurements will (mostly) be deployed by NRENs, while link trace tools will be deployed by both end-sites and NRENs. Even if end users can deploy Ethernet CFM tools on their hosts, that does not mean they can monitor the whole multi-domain Ethernet circuit. There may be a router between the Ethernet circuit provided by the NREN and the end hosts. This means that out-of-band distribution of monitoring data is still required if end-users want to receive Ethernet CFM monitoring information.
3. Transparent transmission
NRENs may provide a transparent Ethernet connection to end-sites through encapsulation of Ethernet frames, thereby obscuring it's topology and intermediate points. This makes it impossible to do monitoring with Ethernet CFM. We recommend that at least the GLIF Open Lightpath Exchanges
(GOLEs) deploy Maintenance Intermediate Points (MIPs) on all circuits, thus revealing part of the lightpath topology.
4. Domain Identifier
The Maintenance Association Identifier (MAID) must be the same for all Maintenance End Points (MEPs) that want to communicate together. Using the VLAN ID of the link as the MAID is impractical with the uptake of VLAN translation. Our recommendation is that all NRENs use the same MAID.
5. MD level
The Maintenance Domain (MD) level need to be agreed upon for both continuity check messages and link trace messages. If the continuity check messages monitor segments between domains, and the link trace monitors the end-to-end connection, the MD level must be different. For example MD level 4 for continuity check messages between domains, and MD level 5 for link trace and loopback messages between end sites.
6. Monitoring Point Identifiers
MEPs should be uniquely identified within a given Ethernet circuit. The MEP Identifier (MEP ID) is an integer in the range 1…8191. This is clearly not enough to assign a globally unique identifier to each MEP. Either the MEP ID must be agreed upon for each individual link, or another identifier should be used. We want to investigate if it is possible to extend the IEEE 802.1ag standard with URN-based globally unique identifiers for each MEP.
7. Time interval between Continuity Checks
The time interval between continuity check messages (CCM) should be agreed upon. A link "fails" after three missed messages. One message every second will result in a link failure detection after 3 seconds. This is fast enough to support the default polling of once every 1 – 5 minutes for out-of-band monitoring with PerfSONAR.

Download file