Re: Approximating Hypothetical Beowulfs

• Messages sorted by: [ date ][ thread ][ subject ][ author ]
• [ Next ][ Previous ] In reply to: Brian Atkins Next in thread: Dan Clemmensen

Brian Atkins writes:

> Yeah besides capacity and architecture (as someone else pointed
> out), I think you might also want to take into account the
> latency of the different network technologies. I think some

You can achieve about ~20 us messaging latency for small messages if you use M-VIA on (DEC Tulip) FastEthernet. Vanilla Linux TCP/IP gives you ~60 us. A cut-through switch introduces a few us overhead. With the channel bonding kernel patch one can coalesce up to 4 NICs into one channel, all of them running full-duplex.

> Beowulfs use special types of networking (non-ethernet?) in
> order to provide faster message passing.

Myrinet, mostly.

• [ Next ][ Previous ] In reply to: Brian Atkins Next in thread: Dan Clemmensen