On 15 Jul 1999, Stuart Lynne wrote:
> How much overhead is there to dragging in the perl just to do the query.
> One of the things I *really* like about exim is the ability for it to
> do 99% of what I need done in a single process with reasonable overhead.
The Apache mod_perl experience says "very little", as long as the perl
interpreter is linked with the binary (dynamically if you like) and the
perl is compiled when the process starts. Apache forks processes (and
doesn't exec) cleverly, so that there's (usually) no significant latency.
The running apache server uses more VM than one without perl, but as it
doesn't exec, and this is all readonly, the forked processes share all
that VM with the parent process.
> I love it and use perl all the time for applications that are not time
> sensitive. But I blanche at the thought of using it for anything I
> need to scale up.
If you are forking, execing and compiling the perl anew to do every menial
task, then no wonder. So don't - Apache demonstrates how to go about this
The Right Way. The Internet Movie Database is probably the best known
really big web server using mod_perl (certainly it started off with
mod_perl, though the current management don't mention underlying software
now).
I appreciate that this is getting off topic for Exim, but the point is
that embedding perl in a server is not to be rejected just because it
"seems big" and is (usually) interpreted; the potential advantages are
great.
Peter Lister P.Lister@??? PGP (RSA): 0xE4D85541
Sychron Ltd http://www.sychron.com PGP (DSS): 0xBC1D7258
1 Cambridge Terrace Voice: +44 1865 200211
Oxford OX1 1UR UK FAX: +44 1865 249666
