Cisco 6509 question

As part of a Citrix performance issue, we are looking at performance on our Cat 6509. We are seeing an issue with high rates of Medium Buffer failures.

When multiple "show buffer" commands are entered, the medium buffer failure count is incrementing from 500 failures a minute to as many as 17000 failures per minute. Medium Buffer misses are also incrementing at about the same rate. The Medium Buffer hit counters increment hardly at all. The Medium Buffer has 0 buffers total, 0 buffers free, 0 permanent, 0 min, and 0 max. It is my understanding that if CEF or dCEF tries to use a buffer and there are none free, it will drop the frame and cause a failure. We tried to add buffers to Medium buffers. However, in "config t" there are no commands for Buffer Medium.

All other buffers appear to be working normal.

First question: Are the failures in the Medium Buffer an issue, especially with performance?

Second Question: Is CEF or dCEF the cause of the high failure rate?

Third Question: Is CEF/dCEF on by default in this version of IOS? There is no command in the config to turn on CEF/dCEF, but CEF/dCEF is running on the switch.

Fourth Question: How do you tune the Medium Buffer if "config T" does not have a command for it.

Fifth Question: Could this be an IOS bug/issue?

Sixth Question: Could this be a hardware bug/issue?

CAT OS 6.3(3A)
IOS 12.1(11)E4

Just to let everyone know. This was an IOS issue we wound up upgrading to the latest stable release and it too the medium buffers completely out of the mix.

Doh ... how on earth did I miss this post ?

I'm dealing with a similar problem at the office pndennie with our backbone catalyst 3550 switches where we are experiencing problems with our VoIP phones - a lot of jittering throughout the network.

While troubleshooting the problem, I found similar issues with our backbone switches where we had quite a few cache misses, but no failures. Further investigation showed that cache misses or fall backs result in the switch processor allocating memory from the public pool buffer - generally not a problem if you have plenty of memory available.

I also contacted Cisco about the problem and they suggested that I upgrade the IOS because of similar problem customers had, which vanished when they upgraded the IOS ... so that's what I'm going to do on Monday morning:)

Cheers,