[clug-talk] Troubleshooting Poor Gigabit Performance
jsc at nddn.net
Thu Mar 29 06:45:15 PDT 2012
It's been a while since I've been on clug-talk. Lots of familiar faces
I'm having some trouble with slow performance and I'm wondering if anyone has
I have two older HP Proliant servers (ML370 and DL360), a more modern desktop
computer and an i7 laptop and I'm having trouble getting any of them to move
data across my network at anywhere above about 45MB/s. While these machines
aren't the pinnacle of modern computing they should be able to saturate GigE.
Server 1 (Goliath) has a NetXtreme BCM5703 Gigabit Ethernet (pciX)
Server 1 (Screamer) has a NetXtreme BCM5701 Gigabit Ethernet (pciX)
Both the switch and ethtool agree that each card is 1Gbit/s full-duplex.
I've tried just piping data from /dev/zero on one machine to /dev/null on the
other and it seems to hit the speed limit.
== Machine 1 ==
06:54:57-root at screamer:~ $ dd if=/dev/zero bs=8k count=1000000| nc goliath 9999
1000000+0 records in
1000000+0 records out
8192000000 bytes (8.2 GB) copied, 167.261 s, 49.0 MB/s
== Machine 2 ==
06:59:19-jsc at goliath:~ $ nc -l 9999 > /dev/null
Strangely (and probably related) is that in the above test, NC pegs a CPU
core on the sending side. I wouldn't have thought that the overhead for
netcat would be significant at all.
I see similar performance (~45MB/s) moving files between machines using Samba.
I've tried with two separate gigE switches (one a new-ish Netgear GS108E) and
a new set of Cat 6 cables and still I get remarkable consistent performance.
Anyone have any ideas what I should be looking at in more detail.
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 836 bytes
Desc: Digital signature
More information about the clug-talk