Alex Rousskov wrote:
>
> Dancer,
>
> I might be in the minority here, but I still suspect that disks are
> irrelevant to your problems.
I suspect that you are correct. The disk subsystem seems to be operating
at one percent of total capacity.
> Here is one more simple test you may want to
> try. Enable "endless" persistent connections on *both* poly-client and
> poly-server using
> --pconn_use_lmt const:5000000 --nagle off
> options, and see if it makes any difference. Try with uncachable replies
> to put disks aside.
Will try it when I get into the office (about 15 mins). Did I give you
the results of that last trial? With uncacheable objects the system
performed about 30 req/sec faster than its best using
rep_cachable:80/dhr:55. Yet physical disk I/O was about the same during
each (virtually nil)
D
Received on Tue Jul 29 2003 - 13:15:57 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:12:06 MST