On Thu, Dec 15, 2005 at 12:55:48PM -0800, Stephen Hemminger ([EMAIL PROTECTED])
wrote:
> Given the pretty extensive modifications necessary, do you have any
> useful benchmark data to show that this is a win. Previous zero
> copy (like page flipping) has not be merged because it did not
> prove to be a net gain.
Yes, I have one.
I use high-performance memory mapped 8139 adapter (original idea
was to reduce CPU usage on my home gateway when downloading huge files
over ftp where I use such an Adapter) on 2.4 Ghz Xeon (1+HT):
1 Gb transfer from 1-way AMD64 machine (2210.767 Ghz), gbit interface, server
uses sendfile()
for data sending.
Client is connected using 8139 Realtek NIC through D-link dgs-1216T Gbit
switch. File is written into ext2 partition.
Speed is the same and is about 10.7-10.9 Mbyte/sec.
CPU usage graph attached.
'zc' - zero-copy transfer.
'recv/write' - recv()/write() transfer.
Tested application attached in first e-mail. It was run with following
parameters:
zc: ./sendfile -a 192.168.0.48 -p 1234 -f /storage1/test -b 192.168.4.78 -B
12345 -t 0 -n64
recv/write: ./sendfile -a 192.168.0.48 -p 1234 -f /storage1/test -b
192.168.4.78 -B 12345 -t 1 -n64
Where only -t differs: it is type of the test.
-n is a number of preallocated pages.
-a - remote addr
-p - remote port
-b - local bind addr
-B - local bind port
-t - 0 - zc, 1 - recv/write test
-n - number of preallocated pages (32 by default).
--
Evgeniy Polyakov
-
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
