On 03/06/16 06:54, Douglas Eadline wrote:
> Many of the 10Gbe (and above) vendors seem to
> be paying more attention to performance than
> before. I like that for my small scale Limulus
> boxes I can get 8-9 usec latency using tcp,
Yeah, a small Haswell cluster on an untuned 100GigE switch that I
All,
https://github.com/gmkurtzer/singularity/releases/tag/2.0
Updated website here:
http://singularity.lbl.gov
--
Doug
--
Mailscanner: Clean
___
Beowulf mailing list, Beowulf@beowulf.org sponsored by Penguin Computing
To change your subscription
> On 02/06/16 05:25, Andrew Piskorski wrote:
>
>> You mean GAMMA (Genoa Active Message MAchine):
>
Used both of these
> GAMMA's last update was 2009, according to their website.
Very good, but only worked for Intel e1000 (IIRC)
> Brice pulled the plug on Open-MX in December 2015.
Worked well,
Yes, I'm pleased that several companies have adopted message rate as a
metric, although I'm also sad that some have gamed the benchmark with
message coalescing.
N1/2 was intended to say something about the shape of the
bandwidth-by-message-size curve. The smaller it is, the farther
you can go with
If anyone is interested in infrastructures using Ethernet like that, have a
look at JASMIN int he UK
http://www.ceda.ac.uk/projects/jasmin/
http://www.channelweb.co.uk/crn-uk/news/2339003/viglen-scents-hpc-success-with-gbp4m-jasmin-deal
Mellanox switching with 10Gbps to the edge and a rather la
