Q: ATM Beowulf

Gerry Creager (N5JXS) gerry.creager at tamu.edu
Wed Nov 12 09:38:10 EST 2003


To expand on that a little bit, the overhead associated withn LANE is 
also going to be a problem for you, even ic communications will be a 
small segment of your cluster activity.

If your application were capable of talking directly at the ATM layer, 
and not have to go through the framing and conversion issues with either 
LANE or IPOA, then you could see some advantages.  However, the issue of 
sending 1500 byte... or 9kB... packets over ethernet, taking advantage 
of the somewhat faster encoding we see for ethernet today, far outstrips 
the potential benefits of ATM on a cluster environment.  And making up 
1500 byte packets, then resending them as 53 byte cells, with 10% 
overhead, just doesn't make sense anymore save in a connection-oriented 
network like WAN or Carrier.  And, fwiw, the carriers are also dropping 
ATM for GBE, 10GBE and carrying same over multiplexed lamdas in the 
glass now.

Gerry

John Hearns wrote:
> On Wed, 12 Nov 2003, Rafael Angel Garcia Leiva wrote:
> 
> 
>>Does make sense to use LAN emulation over ATM for this kind of clusters? Has 
>>anyone experimented with ATM interconnections? Do you think is it 
>>cost-effective today (specially compared to Fast / Gigabit Ethernet)?
> 
> 
> The list will know that a few years ago I was very enthusuastic
> about ATM. I put in a leading edge ATM network at a hospital in
> the UK for medical imaging. I was a proponent of using ATM for clustering.
> 
> These days, I would say the idea is not so good.
> You get built-in Gigabit network interfaces on many motherboards,
> and Gigabit switches are really cheap.
> 
> 
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf

-- 
Gerry Creager -- gerry.creager at tamu.edu
Network Engineering -- AATLT, Texas A&M University
Office: 979.458.4020  FAX: 979.847.8578
Cell: 979.229.5301    Pager: 979.228.0173


_______________________________________________
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