3550xl performace issues

Unanswered Question
Mar 7th, 2007
User Badges:

Hi i've got 26 3550xls and 30 3550smi. the xls are missing buffers big medium and small and the cpu is always running at 60% 70% plus. the smi are fine the run nice and smothly at 5% cpu. Clients on the xls are complaining of slowness. Is there anything i can do to improve this or provide evidence in static that we need new switches. What switches could i recommend, bear in mind the switches need to be able to cascade to each other?

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 5 (2 ratings)
Danilo Dy Wed, 03/07/2007 - 15:47
User Badges:
  • Blue, 1500 points or more

Can you post the "show version", "show proc cpu", and show buffers"?

hoogen_82 Fri, 03/09/2007 - 01:45
User Badges:
  • Silver, 250 points or more

Nothing much you can do. There are no Buffer errors for you to worry about. And as I had told you this document explains you the high CPU utilization http://www.cisco.com/en/US/products/hw/switches/ps607/products_tech_note09186a0080094e78.shtml


Other than that your configuration is good. Minimal you could say. Probably turn off loggin on console, no ip redirects on your vlan interface.

This is what Cisco told me sometime before a Catalyst 2900XL/3500XL with minimal load will

have a CPU utilization between 30% and 80%.



hoogen_82 Wed, 03/07/2007 - 23:36
User Badges:
  • Silver, 250 points or more

Hi do post your outputs inline with what medan has asked for. Along with it do provide your running config.

I believe you are using 3500XL and not 3550XL, there is known issues with it explained in this document. Go through and do what is best for you network, more insight can be provided through your outputs.




network_team Fri, 03/09/2007 - 03:14
User Badges:

From what you have seen within the configure do you think the 3500xl are causing slowness on the network

Danilo Dy Fri, 03/09/2007 - 04:10
User Badges:
  • Blue, 1500 points or more

I think you are using your 3500XLs as access switches where your user are directly connected.

Are you monitoring the bandwidth utilization of the uplink or trunk ports? Avoid daisy-chain in the access-level, it's recommended that access switches are connected to distribution switches and distribution switch should be the root bridge in an STP domain.

Can you post you network topology, showing the connectivities, uplink port speed, STP, switch model, primary and backup root bridge switch?

network_team Fri, 03/09/2007 - 07:23
User Badges:

Yes our configuration is as follows 3 x 3500xl 48 ports daisy chained with two core fibre cables connected to the distribution switches. Uplinks are gigabit fibre and all switch are clients apart from distribution switch and the root is the distribution

Danilo Dy Fri, 03/09/2007 - 07:55
User Badges:
  • Blue, 1500 points or more

Can you save them in JPEG? I don't have VSD at home :(

Danilo Dy Fri, 03/09/2007 - 08:49
User Badges:
  • Blue, 1500 points or more

1. Is there a direct connection between the two distribution switch? Is this connection up?

2. Is the two distribution switch configured as root switches in the STP domain?

3. Have you checked all the switch path to the root switch? Are they taking the shortest and bigger pipe path? No loop?

Danilo Dy Sun, 03/11/2007 - 09:04
User Badges:
  • Blue, 1500 points or more

I just remembered troubleshooting slow performance problem in 2900XL and 3500XL, end up its Cisco production problem.

Check your 3500XL Serial Numbers if they are included in this Field Notice. If they are, contact your vendor. You may not get the same error, but contact your vendor anyway


Other Field Notices for 2900XL and 3500XL.


network_team Mon, 03/12/2007 - 01:53
User Badges:

Yes there is a direct connection between the distribution switches, and yes it is up.

Yes both switches are root STP domain. I shall check the serial numbers of switches

network_team Mon, 03/12/2007 - 05:47
User Badges:

Only 4 of the 26 switch have the hardware fault you are refering to.

Danilo Dy Mon, 03/12/2007 - 06:49
User Badges:
  • Blue, 1500 points or more

Okay, I also checked your IOS version.There's a problem with your current IOS 12.0.5-WC14(ED) up to WC16


Your "show buffer" output is okay, don't worry about it. Your network topology is also okay. High CPU util for 3500XL is normal.

To summarize

1. Contact your vendor to replace the 4 switches affected by the field notice

2. Upgrade 3500XL switches IOS to 12.0.5-WC17(ED)

3. Hardcode all uplinks (including their connection to 4500) to Full-duplex/1000BaseT

Good luck!

network_team Mon, 03/12/2007 - 12:58
User Badges:

Thanks for your help. When say uplinks do you mean the connection from gig module connection from access switch to access switch and then the connection from access switch to distribution.

Danilo Dy Mon, 03/12/2007 - 18:11
User Badges:
  • Blue, 1500 points or more

Yes, all your ports configured with trunking plus the daisy-chain ports. I see in your configuration that the two GE is not harcoded - it's highly recommended that they should be hardcoded.

* access-switch to access-switch

* access-switch to distribution switch


This Discussion