Pekka Pietikainen writes:
> Jumbo frames, sender is dual pIII/500 with 32/66 PCI, receiver a dual pII/450
> with 32/33, both have 1MB Alteons. CPU use is measured using cyclesoak.
> SO_RCVBUF/SO_SNDBUF set to 512k, no other sockopts touched.
Strange, cpu usage is close to nothing for sendfile cases yet full
bandwidth is not obtained. Here is what I am getting on UltraSparc
systems:
ZC results:
2.4.2pre1+zerocopy+sendfile acenic-->Syskonnect 120 MB/sec, 32%
load
2.4.2pre1+zerocopy acenic-->Syskonnect 94 MB/sec, 52%
load
2.4.2pre1+sendfile acenic-->Syskonnect 110 MB/sec, 49%
load
2.4.2pre1 acenic-->Syskonnect 80 MB/sec, 49%
load
I am using the Netgear (ie. 512K memory) acenic cards as the sender.
Your idea to experiment with removing the non-zerocopy related tweaks
from the zerocopy patches may prove fruitful, please let everyone
know what you do or do not discover here.
Later,
David S. Miller
davem@xxxxxxxxxx
|