howdy,
i'm doing some research with the kernel on fragmentation and
performance. we are trying to measure performance of fragmented versus
unfragmented datagrams. nfs seemed like a first place to start, since it
sends fragments in reverse order by default (which we're interested
in) and since it has configurable datagram size. (given some of the
troubles that we're having, these may not be sufficient).
so here's the problem: there do not seem to be datagrams of any
singificant size being sent.
the mount command is:
mount -o rsize=8192,wsize=8192 staff.nm.or:/tmp /mnt/stafftmp
this is the tcpdump output of:
dd if=/dev/urandom of=./tempfile bs=1024 count=500
21:19:44.600273 zapata.nm.org > staff.nm.org: (frag 18330:1296@2960) (ttl
64)
21:19:44.601507 zapata.nm.org > staff.nm.org: (frag 18330:1480@1480+) (ttl
64)
21:19:44.602738 zapata.nm.org.3073878530 > staff.nm.org.nfs: 1472 write
[|nfs] (frag 18330:1480@0+) (ttl 64)
21:19:44.602987 staff.nm.org.nfs > zapata.nm.org.3073878530: reply ok 96
write [|nfs] (ttl 64, id 18876)
21:19:44.620384 zapata.nm.org > staff.nm.org: (frag 18331:1296@2960) (ttl
64)
21:19:44.621615 zapata.nm.org > staff.nm.org: (frag 18331:1480@1480+) (ttl
64)
21:19:44.622847 zapata.nm.org.3090655746 > staff.nm.org.nfs: 1472 write
[|nfs] (frag 18331:1480@0+) (ttl 64)
21:19:44.623095 staff.nm.org.nfs > zapata.nm.org.3090655746: reply ok 96
write [|nfs] (ttl 64, id 18877)
21:19:44.628937 zapata.nm.org > staff.nm.org: (frag 18332:1296@2960) (ttl
64)
21:19:44.630168 zapata.nm.org > staff.nm.org: (frag 18332:1480@1480+) (ttl
64)
21:19:44.631400 zapata.nm.org.3107432962 > staff.nm.org.nfs: 1472 write
[|nfs] (frag 18332:1480@0+) (ttl 64)
21:19:44.631641 staff.nm.org.nfs > zapata.nm.org.3107432962: reply ok 96
write [|nfs] (ttl 64, id 18878)
so, as far as i can tell, the max datagram size is never over 3K, even
though the mount was with 8K (this remains true even when the rsize and
wsize are 64K).
any ideas?
thanks,
todd underwood
todd@unm.edu
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Fri Mar 31 2000 - 21:00:21 EST