[Beowulf] Gigabit Ethernet vs Myrinet

Greg Lindahl lindahl at pathscale.com
Sat Dec 27 12:52:31 EST 2003

On Sat, Dec 27, 2003 at 01:35:04PM +0330, Mahmoud Saleh wrote:

> Reading a couple of comparison tables regarding latency of different NIC 
> protocols, I noticed that many solutions suggest to use Myrinet style 
> NIC due to its low latency, namely around 8usec for I/O intensive 
> cluster jobs. I was wondering if Gigabit Ethernet does the same.

First off, most people separate disk I/O from program
communications. I'll assume that you're talking about the second.

> Suppose that Maximum packet size in GE is 1500 bytes and the minimum is 
> aroud 100 bytes. This translates to an average of 800 bytes or 6400 
> bits. In Gigabit Ethernet that would cause a delay of 6400/10^9 sec  or 
> 6.4usec for packet assembly, which is in the same order as Myrinet. 

You are only thinking about the time needed to send the actual
bytes. The total time to send a small message is much bigger than
that. There are published papers that show the "ping pong" latency for
gigabit ethernet. This number is highly dependent on the exact
gigE card, switch, OS, and gigE driver that you're using.

-- greg

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