networking options

Håkon Bugge Hakon.Bugge at scali.com
Tue Sep 24 18:04:24 EDT 2002


At 05:28 PM 9/24/02 -0400, beowulf-request at beowulf.org wrote:
>Date: Tue, 24 Sep 2002 11:42:07 -0500
>From: Richard Walsh <rbw at ahpcrc.org>
>To: dtj at uberh4x0r.org, mp00aa at cosc.brocku.ca
>Subject: Re: networking options
>Cc: beowulf at beowulf.org
>
>
>Matt Pratola wrote:
>
> >Of course each app is different, but i'm interested more in high
> >throughput/low latency as it cannot harm apps which don't require this,
> >but is absolutely necessary for apps which do require this level of
> >performance.  In this regard, myrinet is certainly the one i've heard the
> >most about, but that doesn't preclude something else which is new and/or
> >better.
>
>OK. Your most obvious choices in 2002 with rough performance estimates
>(these are generally better-than MPI numbers) are:
>
>        Technology               ~Best Case Bandwidth    ~Hardware Latency
>
>     1. Fast Ethernet*              ~10 Mbytes/sec       ~50-75  usecs
>     2. Channel Bonded FE        ~20-25 Mbytes/sec       ~50-75  usecs
>     3. Gigabit Ethernet*        ~60-110 Mbytes/sec      ~50-200 usecs
>     4. Channel Bonded GE        ~????? Mbytes/sec       ~50-200 usecs
>     5. SCI                      ~150-200 Mbytes/sec     ~4-8    usecs

Since you use the term best-case, the numbers for SCI are somewhat better; 
around 260MB/s for e7500, 280MB/s for AMD 760 MPX, 320 for Itanium1, and 
375 MB/s for HP zx1 Itanium2.

Håkon

>     6. Myrinet                  ~200-250 Mbytes/sec     ~7-10   usecs
>     7. Quadrics**               ~300 Mbytes/sec         ~4-5    usecs
>
>     8. Cray T3E                 ~200-400 Mbytes/sec     ~1-2    usecs
>
>     *  GAMMA MPI can improve latencies signficantly.
>     ** Quadics claims 700 Mbytes/sec in 2003 (dual rail?)
>
>Does not include less convention/available things like Infiniband or
>Fiber Channel.
>
>These data are from memory and subject to varification and refinement.
>You can check on proce-performance and scaling issues.
>
>Hope this helps,
>
>rbw

--
Håkon Bugge; VP Product Development; Scali AS;
mailto:hob at scali.no; http://www.scali.com; fax: +47 22 62 89 51;
Voice: +47 22 62 89 50; Cellular (Europe+US): +47 924 84 514;
Visiting Addr: Olaf Helsets vei 6, Bogerud, N-0621 Oslo, Norway;
Mail Addr:  Scali AS, Postboks 150, Oppsal, N-0619  Oslo, Norway;

_______________________________________________
Beowulf mailing list, Beowulf at beowulf.org
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf



More information about the Beowulf mailing list