EIGRP Delay Mismatch Between SPA Card and Remote

[akrauska]

We have a DS3 Spa 2XT3/E3 in 6509/mpls router to handle frame-relay to 130 remote routers. Spa card serial and sub-interfaces have a delay of 200 usec and remotes have a delay of 20,000 usec. Is this a delay mis-match? It seems to work but the route metrics from hub to spoke and vice-versa differ by over 500,000 units.

Is this correct or should we set a manual metric on each sub-interface of 20,000 to match the T-1 (1544) for that individual connection. TAC told me the the delay should only be set on the serial interface, not the sub-interfaces. Also, all 130 remotes are coming into a single serial interface, though half are not full T-1, but fractional. Though the uitilization is ok, I am seeing input queue drops throughout the day and am wondering about the impact of having 130 routers going through this single DS3 serial interface. You comments are appreciated. Thanks.

[Giuseppe Larosa]

Hello Andrew,

the delay mismatch is not an issue until each remote router has only one FR PVC to the hub router.

In order to deal with multiple remote routers you can :

a) increase the input hold queue to its max value

b) you can play with the bandwidth command and the ip eigrp bandwidth-percent in order to shape the eigrp traffic on each FR PVC : the idea is to use a low bandwidth and then to provide space for eigrp traffic with the second command

The questions you need to address are the following :

Is the access link at the hub site oversuscribed ? Ds3 and 130 T1 /subT1

If it is use as bandwidth = access link speed/# DLCI

In addition if you haven't already done configure all remote routers as eigrp stub routers if possible (if no other routers are behind them).

This will save bandwidth during an EIGRP query.

Hope to help

Giuseppe