[Catalyst] Resolving performance issues.

[Len Jaffe]

--- Bill Moseley <moseley at hank.org> wrote:

> On Tue, Apr 25, 2006 at 12:21:53PM -0700, Len Jaffe
> wrote:
> > You need a way to take the network latency out of
> your
> > tests.  One admittedly dumb way is to run
> traceroute
> > while running yur tests, and subtract the network
> > latency. I said it was dumb.
> 
> Both tests were done on the local machines using the
> Catalyst test
> server connecting to localhost.  So probably not
> that much latency
> involved.
> 
> Both with up-to-date versions of the application and
> Catalyst.  May be some differences is other modules,
> as it's
> perl 5.8.4 compared to my dev machine 5.8.8.

The accept() call was an order of magnitute slower at
dream host.  Send() was also slower.  Right off the
bat, there is ahuge difference in the speed of the
TCP/IP stuff.  

It could be attributable to differences in OS, TCP/IP
stacks implementations/patches, etc.

Len.