G'day,<br>
<br>
I need help optimizing a server so it can maximize incoming network traffic (UDP) and <br>
write the data to disks over a 100T network connection. My system works well with <br>
no dropped packets until I turn on disk writes to the SATA disks. I have tried tuning <br>
the network parameters but with marginal success. <br>
<br>
Motherboard/CPU: Supermicro with dual 1000T interfaces, TG3 driver, 3.2G P4 (HT enabled)<br>
Disks: dual 80G SATA disks<br>
Network: TG3 driver<br>
O/S: Fedora Core 4, 2.6.17-1.2142<br>
<br>
Tuning options used:<br>
via kernel: /proc/sys/net/core/rmem_max 2500000<br>
/proc/sys/net/core/wmem_max 2500000<br>
/proc/sys/net/core/netdev_max_backlog 20000<br>
via ethtool, interrupt coalesce itmes:<br>
rx-usec 100<br>
rx-frames 20<br>
via ethtool, RX ring size to max allowd by driver:<br>
rx 511<br>
<br>
Disk:<br>
originally SW RAID, experienced lots of errors when write was enabled<br>
changed to single partition, formatted with mke2fs -m 0 -b 4096, far fewer write errors<br>
<br>
I am still getting groups of dropped packets when cron.daily or weekly runs and at random times.<br>
The errors seem to be groups of missed packets and rarely a single packet. I did find a similar <br>
problem that someone else reported and am wondering if this could be the same problem <br>
(both are using Supermicro motherboards and the on-board TG3 driver):<br>
<a href="http://marc.theaimsgroup.com/?l=linux-ide&m=115689223417994&w=2">http://marc.theaimsgroup.com/?l=linux-ide&m=115689223417994&w=2</a><br>
<br>
Any ideas? Can I lower the priority of writes to disk? Are there serial options I can use? Are<br>
there updates to the tg3 driver I could try (for example really large rx buffers beyond 511)?<br>
<br>
Cheers,<br>
--<br>
Wade Hampton<br>