1800/24 G low performance.

[Diego Castelli]

Hi all,
a client complains low perf. of the switches (he has 3).

either if a single cable is connected to the switch the ping time to the management interface is always more than 4 ms.

the first two packets' time is more than 400 ms....

what can i do?

[Pieter 't Hart]

at first
4ms is not alarming.
the lowest value displayed is 1 ms.
but i agree with a dirrect to the destination you should expect a stable low value

second
a little more info will help.
whats the ip-configuration of the client doing the ping,
whats the network config of the destination switch and switches in between.
what routing is configured between those networks.

apart from this slow respons to ping-times do not neccessarily map to a performance problem.

third
for the first packets the network first must do a "ARP request" that is a query on the network to determine anyone who knows the mac-adress of the host having a certain ip-adress.
if this is not on the local subnet, then multiple queries are neccessary before a path to the destination is determined.
only then the real ping can do it's work.
so it's normal the first ping respons takes longer then subsequent responses.

[Diego Castelli]

Hi Peter,

First of all thanks for the reply.

I agree with you about the first ping taking longer... Thanks for info!

I analyzed performances with solar winds Network performances tool and it gave me very bad statistics about the response time (with very low packet loss). The statistics of another HP switch in the same LAN but another model (i don't remember yet) are good (1ms ant very low resp. time).


SOME DATA

The switches are SAME LAN, no VLAN configurations, Fiber backbone between them.
The IP Configuration is C Class 192.168.1.x
The Host doing the ping is 2 switches far (but it's 3 switches far from the other good responding switch).

Looking at the Web interface i didn't get useful info.

Thanks again for your answer and future info.
Diego

[Matt Hobbs · Accepted solution]

Don't ping the switch as a measure of performance. That is simply testing the management interface of the switch which is often runs at a very low priority and is completely seperate from the switching function.

If doing any ping tests, ping between differents on the same switch. That will give you a much better result.

By chance, this poor performance the client is complaining of, does it involve any 100Mbit devices? The 1800 is gigabit optimised so will always perform much better with gigabit clients. When it has to deal with 100Mbit devices quite often the internal port buffers deplete too quickly. You can try enabling flow-control on those ports, but it would be a better idea to upgrade the NIC if possible.

[Diego Castelli]

Thank You Matt.

Poor performance are present accessing a simple host too.

There are more than one 100Mbit device attached to those switch.
But why the same device should perform better when attached to a "non Gigabit-optimized" switch? Is there a way to resolve this?

Could "enabling flow control" be a resolution?
What config do you recommend?

Thanks in advance.
Diego

[Matt Hobbs]

From my experience, if your clients are 100Mbit and are copying files from a Gigabit server, this is when the 1800 will suffer the most.

Enabling flow-control on all ports and all devices connected to the switch should help to a certain degree.

There are also some registry edits you may be able to make to the 100Mbit client machines. Check out "TCP Optimizer".