High Packet Loss with Catalyst 2950

Jan 19, 2006 4 Replies
P

I have a client that has a Cisco Catalyst 2950 48 port switch with two



1000-T interfaces. They are experiencing slow access speeds for their servers, so I checked the port on the switch, it showed that the port for their main server was getting extremely high error rates (79,000 over 24 hours when I reset it) I tested the wiring, it appears to be good, I have pretty much eliminated everything except the switch, so I decided to pull that out of the network, putting back in their old Linksys switch with another switch to give them the ports they needed, and brought the Catalyst back to my office.

Yesterday my client called me and reported that they are noticing that they screens in their applications are pulling up a little bit more quickly.



So it's up to me to find whats causing the issue, and fix it, unfortunately, I can't makes heads or tails of it, which would a cheap Linksys switch give them better performance, when the Cisco switch seems to be well setup (I didn't do the setup, just checked it out).



They have a single subnet, no VLANs, or anything else that I can think to cause issues. This was their core switch going out to all the computers, they use about 30 of the ports.



Shawn



Vote
Twitter Facebook LinkedIn
S

Sounds like your speed and duplex settings are mismatched. Try hard-coding the speed and duplex on the server and switchport (100/full duplex). If it is indeed a port speed/duplex mismatch, note whether you're seeing the "late collisions" counter increment. If you *know* that your cabling is in good shape and in spec, then this counter should never increment...it could if hardware is bad, but it's still likely a port mismatch.

Good Luck!

Vote
Twitter Facebook LinkedIn
K

First thing that comes to mind for some reason is the speed/duplex settings of your Catalyst switch. Are they hard coded to some speed and duplex on the Cisco? Chances are that the Linksys is set for auto-negotiate and that means the hosts probably are too. If your Cisco is hard-coded to 100 full lets say, then your hosts will end up negotiating something funky like 10-half or something.

Also, what kind of errors were they? That will go a long way to help determine the problem. Not all errors are the same.

Vote
Twitter Facebook LinkedIn
P

It was just generic packet errors reported by the switch web console. I am going to try reinstalling it with the speed hard coded for that particular server.

Vote
Twitter Facebook LinkedIn
W

I had two 2950 switches at a remote office location and I think they stink. They are connected back to my home office over a fiber-optic link and a pair of Canary 100BaseFX media converters. I did good to get

40Mbps true thruput thru the 2950's. No duplex mismatches, just high error rates and overall lackluster performance. I replaced one with a cheap Linksys consumer-grade layer2 switch and another with an old 3Com 10/100 hub and got back up to 60+ Mbps sustainable thruput from a host at one end to a host at the far other end. I'll never buy a 2950 ever again.
Vote
Twitter Facebook LinkedIn

Join the Discussion

Have something to add? Share your thoughts — no account required.

Didn't find your answer?

Ask the community — no account required