sorry ... not cut through
lindahl at keyresearch.com
Wed May 15 10:38:28 PDT 2002
On Tue, May 14, 2002 at 12:18:56PM -0400, Robert G. Brown wrote:
> Regarding cutthrough vs store-and-forward -- some years ago this issue
> was discussed on the list and I grew curious as to who made cutthrough
> switches, which seemed likely to do better on bandwidth if perhaps worse
> on latency.
You said this backwards: cutthrough switches ought to be better on
latency. What it means is that the switch doesn't have to receive the
entire packet before it starts sending it out the other side. Store
and forward does exactly what it says: it stores the entire packet
before it thinks about forwarding it.
At 10 mbps this is a Really Big Deal: a 1500 byte packet takes 1.5
milliseconds to receive a packet, so that's a huge difference in
At 100 mbps that's 150 microseconds, which is still noticable using
At 1 gigabit, it's only 15 microseconds, and only people using
protocols like M-VIA could notice. I suspect this is why most (all?)
smaller switches don't do cut-through.
Cut-through vs. store-and-forward ought not affect bandwidth at all.
More information about the Beowulf