Re: Squid vs. NetCache

> Hmmm, speed the cache up by giving it more to do... dunno about that ;)

        But it works!

        The other quick fix alternative is to change the MAX_POLL_TIME
definition in comm.c from 50 down to say 10. This burns a little more CPU
by making it poll the async threads 5 times faster (under low loads). This
should be good for about 400 KB/sec (single large web object throughput)
with the sort of hardware most people use. In fact, your minimum CPU usage
will go up from .01% to .05% (measured on Solaris 2.5.1 167 MHz CPU). If
that doesn't bother you (as I imagine it wouldn't) then this change might
be worthwhile if you want even faster low-load squid-async performance.

        Cheers,

                Stew.

P.S. George, I think just polling it faster it probably the easiest approach
and with the sort of CPU difference we're talking about (very little to not
much) more advanced mechanisms probably aren't worthwhile. It still isn't a
dynamic approach but we have to look at the end-result and it's going to be
acceptable for practically everyone.
Received on Thu May 28 1998 - 19:29:56 MDT