cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
510
Views
0
Helpful
4
Replies

Unusual response time from long-term pinging on LAN

packet919
Level 1
Level 1

Has anyone seen any issues where pinging to any point on the LAN will show an occasional slow reponse (anywhere between 80-400 ms) at random intervals (anywhere between 1 out of 50 to 1 out of 200) and under low-load conditions? This occurs even when pinging from one switch to another switch that is directly connected, or when pinging to directly connected hosts. We're in a mixed-vendor switched environment (we're migrating away from Extreme to Cisco), and I'm tearing my hair out trying to find out if there's a real issue or if this is something inherent to our design or to the gear we're running. No misconfigured NICs around, no chatty or unusual broadcast/multicast machines or traffic, peak load on edge-to-core links is 12-15%, 48-port edge switches aren't even half-full. I've run a Fluke tester on it, I've done protocol analysis, and I've yet to find the cause. I've never really noticed anything like this on any network I've worked on. Has anyone seen anything like this before???

4 Replies 4

mkaroly
Level 1
Level 1

Hi

wich version of IOS did you run on your switches? Did yoy try to modify the speed on port and on NIC's ?

Karoly

rfroom
Cisco Employee
Cisco Employee

Are you pinging from or to the switch IP address or a router IP address. ICMP packets to the switch managment interface is given the lowest priority and the response time will vary dramatically. I strongly suggest pinging between 2 workstations if not already doing so.

What switches do you have and what software versions.

Check to see if you have multicast/broadcasting getting broadcast every so often and check STP TCNs.

In the case of the Cisco switches, yes, I'm pinging to the sc0 address. We have pinged between hosts as well as between switches, and the results are very similar. Our Cisco edge switches are 3548s, running 12.0(5.2)XU. STP is not running, broadcasts/multicasts on these segments is very low (2-5%), and there are no errors. We have confirmed all this with our Fluke LANMeter. There are no obvious issues with speed/duplex mismatches, no obvious switch resource issues, and no obvious problem traffic generated from hosts on any segments (i.e., bad/misconfigured NIC, overly chatty protocol problems, etc.). The only obvious difference on a couple of the affected segments is a much higher-than-usual amount of fragmented IP traffic (due to NFS), but segments with none or low amounts seem to have the same issues.

When you ping the sc0 interface of CatOS switch or the VLAN IP address of XL switch you will definitely see varied ping response times in the range you indicated. This is normal behavior as the process gives the lowest priority to ICMP traffic directed to it. In addition, ping the sc0 interface or VLAN interface is in no way a measure of performance. Catalyst switches do traffic forwarding in hardware and pinging the sc0 interface or the VLAN ip address is not a measure of performance whatsoever.

Therefore, for all tests, ping between hosts. You mention you are using NFS. NFS is very bandwidth extensive and if the server is attached gigabit, you may overrun the output buffer. In other words, trying to send more than 100 meg of traffic out a 100 meg port. Look at the show mac on the CatOS switches to see if you see any outlost. The show interface on the XL switches will indicate output buffer drops or something similiar. Also, see if you are buffer starving the XL, check out the following link for info on how to do that: www.cisco.com/warp/customer/473/19.html.

Please tell me more about the core switch?Is a Cat6k? What version of code and what line cards?

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: