TCP/UDP cksum offload on hme(4)
Tillman Hodgson
tillman at seekingfire.com
Thu Jun 17 14:11:11 GMT 2004
On Thu, Jun 17, 2004 at 03:49:35PM +0900, Pyun YongHyeon wrote:
> I can't saturate the wire with/without my patch.(For TCP without
> witness it was about 67Mbps. On OpenBSD I got about 93Mbps.)
> I vaguely guess the reason was caused by hme's lack of
> auto-negotiation.
That seems odd to me ... on my Ultra 5 (with 5 hme ports) running
-current as of April 16, I recently ran some IPsec vs clear transmission
tests:
Ultra 5, clear transmission:
[root at caliban /usr/local/netperf]# ./netperf -t UDP_STREAM -H athena
UDP UNIDIRECTIONAL SEND TEST to athena : histogram
Socket Message Elapsed Messages
Size Size Time Okay Errors Throughput
bytes bytes secs # # 10^6bits/sec
9216 9216 10.01 13004 13160 95.81
42080 10.01 12778 94.14
Ultra 5, with 3des:
[root at caliban /usr/local/netperf]# ./netperf -t UDP_STREAM -H secathena
UDP UNIDIRECTIONAL SEND TEST to secathena : histogram
Socket Message Elapsed Messages
Size Size Time Okay Errors Throughput
bytes bytes secs # # 10^6bits/sec
9216 9216 10.01 715 0 5.27
42080 10.01 713 5.25
Ultra 5n, with blowfish:
[root at caliban ~]# /usr/local/netperf/netperf -t UDP_STREAM -H secathena
UDP UNIDIRECTIONAL SEND TEST to secathena : histogram
Socket Message Elapsed Messages
Size Size Time Okay Errors Throughput
bytes bytes secs # # 10^6bits/sec
9216 9216 10.01 14744 0 108.63
42080 10.01 3681 27.12
(108, greater than 100mbit max, is due to using 'deflate' compression
algorithm)
(Oh, and the moral of the story is that blowfish rocks when usign IPsec
in transport mode :-))
This was testing a UDP stream (simulating my UDP NFS exports) to an x86
NFS file server runing -STABLE with the a gigabit interface. They were
connected through a D-link 1026G switch, and auto-negotiation was used.
-T
--
Zen is the unsymbolization of the world.
R.H. Blyth
More information about the freebsd-sparc64
mailing list