On 01/23/2015 02:39 PM, Bill Wichser wrote:
We had a strange event last night. Our IB fabric started demonstrating some
odd routing behavior over IB.
host A could ping both B and C, yet B and C could not ping one another. This
was only at the IP layer. ibping tests all worked fine. A few r
Hi Mark
On 01/23/2015 07:36 AM, Mark Hahn wrote:
> It seems like 10gT is on the cusp of real volume-type prices: I saw a
> quote today for a major-vendor 24pt switch for something like $140/port
> after educational discounts. I know 10gT is somewhat higher latency
> than other PHYs - or is it wo
On 12:41PM Fri 01/23/15 -0500, Joe Landman wrote:
> Interesting.
I'm taking my ball and going home for today. Thanks for the info.
--
Gavin W. Burris
Senior Project Leader for Research Computing
The Wharton School
University of Pennsylvania
___
Beowu
On 01/23/2015 12:14 PM, Gavin W. Burris wrote:
On 12:04PM Fri 01/23/15 -0500, Joe Landman wrote:
Don't use ping for HPC latency/throughput tests.
What is your preferred benchmark, Joe?
Interesting.
Ping isn't a benchmark. Its a very basic test of connectivity, packet
transfer and routin
On 12:04PM Fri 01/23/15 -0500, Joe Landman wrote:
> Don't use ping for HPC latency/throughput tests.
What is your preferred benchmark, Joe?
--
Gavin W. Burris
Senior Project Leader for Research Computing
The Wharton School
University of Pennsylvania
_
On 01/23/2015 12:03 PM, Gavin W. Burris wrote:
min packet ping-pong latency is the standard metric, though packets/sec
rate might be interesting as well.
# ping -q -s 1 -f hpcc001-i -c 100
100 packets transmitted, 100 received, 0% packet loss, time 47132ms
ipg/ewma 0.047
> min packet ping-pong latency is the standard metric, though packets/sec
> rate might be interesting as well.
# ping -q -s 1 -f hpcc001-i -c 100
100 packets transmitted, 100 received, 0% packet loss, time 47132ms
ipg/ewma 0.047/0.000 ms
100/47132 = 21.217 packets/ms
Go
Hi, Mark.
Here are some tests, Let me know if you'd like something specific.
# lspci |grep Ether
01:00.0 Ethernet controller: Intel Corporation 82599 10 Gigabit Dual Port
Backplane Connection (rev 01)
01:00.1 Ethernet controller: Intel Corporation 82599 10 Gigabit Dual Port
Backplane C
Mark,
I have been pleasantly surprised by the performance of Chelsio.
My tests are very small scale (no switch) and the hardware I'm
testing is for my Limulus desk side cluster.
I ran netpipe TCP and got latencies of 9us back-to-back (SFP+)
Hopping through a two port adapter I got 18 us.
I the
We have been using 10g CX4 for the past 7yrs for our cluster core
network and are just going fully into 10gT. But our focus is more on
bandwidth/HTC instead of HPC, i.e. we do not care so much about latencies.
I'm sorry if I was unclear: I'm interested specifically in whether
recent development
On 01/23/2015 01:36 AM, Mark Hahn wrote:
Hi all,
I'd appreciate any comments about the state of 10G as a reasonable
cluster network. Have you done any recent work on 10G performance?
https://lwn.net/Articles/629155/
shows some excellent evidence-based work on kernel paths, but it seems
focus
Thanks for the suggestion. Yes, we have been through that. Nothing new
though not clean either. But the issues complained about in the file
have been there for some time so I don't believe this file sheds any
light. Of course watching this as the switch infrastructure returned
was quite noi
Do you see anything interesting in the opensm logs on that server?
In the past I have found looking through opensm logs to be tough going though -
generally full fo verbose messages which don't mean a lot.
Maybe if you could track down the time when the problem first started and look
in the ope
We had a strange event last night. Our IB fabric started demonstrating
some odd routing behavior over IB.
host A could ping both B and C, yet B and C could not ping one another.
This was only at the IP layer. ibping tests all worked fine. A few
runs of ibdiagnet produced all the switches a
Hi, Mark.
We just launched a new cluster, all with 10gigE. What kind of numbers
are you looking for? Most of our workload is data analysis or
iterations of a simulation with commercial software, though we do
validate and support Open MPI. We just opened the doors and are still
working the kinks
It occurred to me the other day that it's about time we had something better
than 1GE for commodity networking. It's good news that switch costs are coming
down but I've yet to see a server with an onboard 10gT adaptor (although I have
seen some with SFP+ 10g).
Supermicro systems are available
On Fri, 2015-01-23 at 01:36 -0500, Mark Hahn wrote:
> Hi all,
> I'd appreciate any comments about the state of 10G as a reasonable
> cluster network. Have you done any recent work on 10G performance?
>
> https://lwn.net/Articles/629155/
I had a go at using RoCE with some Mellanox NICs a year or
17 matches
Mail list logo
