ICMP benchmark?

Discussions of the Ethernet/IEEE 802.3 protocols.

ICMP benchmark?

Postby Dan Stromberg » Tue Nov 22, 2005 3:11 am

Hi folks.

We (at UCI) have a gigabit network we need to troubleshoot. It seems to
be underperforming quite a lot - like it's performing about 1/10th of what
it should.

Some of the routers and switches aren't under our control. We have some
SNMP data getting graphed for our router, but getting that same data for
the other equipment could (or maybe not!) hit bureaucracy.

Which leads me to ask - is there such a thing as a network benchmark that
doesn't require the usual client-server architecture, like maybe one
that's based on ICMP rather than TCP or UDP?

If there is such a thing, and it gives results that are somehow related to
what we might expect from TCP and/or UDP, then we might be able to run it
against each router hop, to see at what point things are slowing down,
without having to have router access.

PS: I've already tried pchar, clink, pathchar and pipechar. On this
network, pipechar seems to come the closest to giving a useful result, but
I'm not sure I trust it, since the numbers it gives vary so much from run
to run, and because it isn't showing a drop off point - it thinks the
network is much faster than it is, even faster than what iperf thinks.

Thanks!
Dan Stromberg
 

Re: ICMP benchmark?

Postby Robert Redelmeier » Mon Nov 28, 2005 12:20 pm

In comp.dcom.lans.ethernet Dan Stromberg <strombrg@dcs.nac.uci.edu> wrote:
We (at UCI) have a gigabit network we need to troubleshoot.
It seems to be underperforming quite a lot - like it's
performing about 1/10th of what it should.

For quick troubleshooting I use `ttcp`. But for saturating
gigabit, you need very fast machines and drivers -- check by
running against localhost.

Which leads me to ask - is there such a thing as a network
benchmark that doesn't require the usual client-server
architecture, like maybe one that's based on ICMP rather
than TCP or UDP?

Certainly! One of the easiest bandwidth tests is using
`ping`, usually with larger packet and flood options.

-- Robert
Robert Redelmeier
 

Re: ICMP benchmark?

Postby Joe Beanfish » Mon Nov 28, 2005 8:02 pm

On Mon, 28 Nov 2005 10:01:38 -0500, Robert Redelmeier <redelm@ev1.net.invalid> wrote:

In comp.dcom.lans.ethernet Dan Stromberg <strombrg@dcs.nac.uci.edu> wrote:
We (at UCI) have a gigabit network we need to troubleshoot.
It seems to be underperforming quite a lot - like it's
performing about 1/10th of what it should.

Ensure that BOTH ends of the connections (your computer and the switch it's
plugged into) are using full-duplex.
Joe Beanfish
 

Re: ICMP benchmark?

Postby bigoldjimbo » Thu Dec 01, 2005 7:45 pm

Could always try Ixia QCheck. as long as both ends are running this
software you can evaluate throughput among other things to try to narrow
down the problem.


Joe Beanfish wrote:
On Mon, 28 Nov 2005 10:01:38 -0500, Robert Redelmeier
redelm@ev1.net.invalid> wrote:

In comp.dcom.lans.ethernet Dan Stromberg <strombrg@dcs.nac.uci.edu
wrote:

We (at UCI) have a gigabit network we need to troubleshoot.
It seems to be underperforming quite a lot - like it's
performing about 1/10th of what it should.


Ensure that BOTH ends of the connections (your computer and the switch it's
plugged into) are using full-duplex.
bigoldjimbo
 


Return to Ethernet

Who is online

Users browsing this forum: No registered users and 0 guests

cron