Respone time unstable

[YenLin]

Hi all,

we've got 2 model or procurve switches.

Both of them have them same situation.

I use my pc to ping them, it would get a low respone time in every 5-6 times.

Their respone time shows as below:

Reply from 192.168.4.205: bytes=32 time=6ms TTL=255
Reply from 192.168.4.205: bytes=32 time=1ms TTL=255
Reply from 192.168.4.205: bytes=32 time=1ms TTL=255
Reply from 192.168.4.205: bytes=32 time=1ms TTL=255
Reply from 192.168.4.205: bytes=32 time=1ms TTL=255
Reply from 192.168.4.205: bytes=32 time=28ms TTL=255
Reply from 192.168.4.205: bytes=32 time=1ms TTL=255
Reply from 192.168.4.205: bytes=32 time=1ms TTL=255
Reply from 192.168.4.205: bytes=32 time=1ms TTL=255
Reply from 192.168.4.205: bytes=32 time=1ms TTL=255
Reply from 192.168.4.205: bytes=32 time=66ms TTL=255

HP have been saying icmp is not a good way to identify.

They said it can caused by high traffic, cpu or low memory.

But, our CPU and memory are around 4% and 20%, respectly.

Another answer is related to QOS.

But, I don't believe so.

The models are J9627A-2620-Poe switch and J9138A-2520-Poe switch.

2620 flash version is: RA.15.10.0010

2520 flash version is:S.15.09.0012

Both of them are latest verion that HP announced.

Any suggestion to fix it?

[Vince_Whirlwind]

It's true that ICMP isn't a good way to measure performance, because the ICMP is given a very low priority.

However, that looks like a very interesting pattern of latency hiccups, and I would be as curious as you are.

The next thing I might try to satisfy my curiosity would be to setup a large transfer on the same path, and to do a packet capture to see if anything odd is happening. I would do the same data transfer between the hosts but using a direct cable for a comparison.

[YenLin]

It's interesting that we've got PIM DM in our environment.

But, we don't have PIM DM at our Sydney office, but they also have same symton.

Actually, our network is quite old and our major servers, hosts, printers are all located at vlan 1.

But, I don't see any reason why they stay at vlan 1 would have this kind of status.

For example, if one switch only connect to my NB.

It still has same situation.

What do you think?

There isn't any odd or unneccessary packet within wireshark capture.

BTW, our 2520-8 and 2530-POE don't have this kind of problem.

Neither as non POE switches.