cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
377
Views
0
Helpful
6
Replies

slow performance

vkrishnan
Level 1
Level 1

We run a gigabit backbone at two sites. The core switch at one site is a cisco 3550-12T (1 server & 70 workstations) and the other an 3550-12G(15 servers and 400 workstations). Performance after migrating to them has been extremely slow (previously used cisco 1900 series switches and performance was better with them!!). All other switches are cisco 3548xl's and 3550-24 or 3550-48. We currently use both IP and IPX/SPX protocols. Performance for IPX/SPX connections is significantly better than for IP. Any tools to monitor performance of these core switches might help or any methods to follow that might lead to a clue would be appreciated.

6 Replies 6

efrahim
Level 4
Level 4

What sofware version you are using on both 3550- 3548.

do sh interface to make sure you are not seeing any errors

There appear to be no collisions or errors. The IOS versions are:

IOS (tm) C3550 Software (C3550-I5Q3L2-M), Version 12.1(8)EA1c, RELEASE SOFTWARE (fc1)

IOS (tm) C3500XL Software (C3500XL-C3H2S-M), Version 12.0(5.3)WC(1), MAINTENANCE INTERIM SOFTWARE

Erick Bergquist
Level 6
Level 6

Are you routing between multiple IP networks (VLANs) or is everyone in the same VLAN?

Is IP traffic slow going between the sites or between users at the same site?

Only 1 VLAN and traffic is slow between users and servers located at the same site. Two sites are connected via a fractional T1 and traffic is limited to AS/400 ip traffic and email.

Earlier had mentioned there were no collisions or errors and this applies to ports where the other 3500 series are connected. There are collisions on ports where a router and hub are connected to the backbone switch and on ports connecting workstations though.

Collisions are to be expected (and are quite normal) on half-duplex interfaces (such as ones that connect to a hub, & depending upon what series.. a router too.) I would recommend checking for duplex mismatches & hardcoding the duplex/speed on each port/interface, instead of having it auto-negotiate. I would also check to make sure there are no spanning-tree loops going on in the network (such as having the portfast feature configured on ports that are directly connected to another switch, router, or anything that can generate BPDU frames.) A good way to prevent against having loops like that inadvertently happen, is to implement errdisable traps (bpdu-guard, etc.) How many users do you have sharing this 1 VLAN? Just my 2 cents.. hope this helps.

- Matt

One site has approx. 550 nodes and the second site has approx 75 nodes. We have checked to make sure any ports that connect to other switches do not have port fast enabled. Speed/duplex has been forced on the ports to match nics on nodes connected. We have looked at bandwidth stats on the backbone switches and others in each closet. No link is using more than 1.5% (unless I am reading this wrong). The average is closer to 0.08xx% on these ports.

Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community:

Innovations in Cisco Full Stack Observability - A new webinar from Cisco