Someone wrote to me privately and said:
> The first block causes the remote system to send an ARP anoucement.
I don't think so:
1. I generally run several experiments so that, among other things,
the ARP caches are warm.
2. The gap in throughput seems to be independent of the amount of
data transferred. If the difference were due to a start-up op
like ARP, the gap should drop in larger transfers. But it does
not:
Transfer Ordinary 4 byte exchange
Size (Mb/s) (Mb/s)
8MB 52 57
128MB 52 57
500MB 52 57
3. There are no ARPs in the tcpdumps.
Vic
-
: send the line "unsubscribe linux-net" in
the body of a message to majordomo@vger.kernel.org
