----- Original Message ----- From: "Davide Libenzi" <[EMAIL PROTECTED]> Newsgroups: saltstorm.xmail Sent: Friday, November 22, 2002 8:15 PM Subject: [xmail] Re: outgoing filters
> > On Fri, 22 Nov 2002, Newsmirror wrote: > > > The reason Rob initiated this thread was based on a discussion Rob and I had some > > days ago about how to further kill off the remaining bottleneck in a Scope-like > > client/server filter process, which is spawning the client. Having a ~5kb client > > binary in the middle might not be big deal performancewise if run once every 5 min, > > but on a heavily loaded XMail server loading it every other second 24-7 makes it a > > bottleneck in the long run, thus the idea came up it would be nice to make the >rather > > trivial client code part the XMail codebase somehow. > > > > I think such an addition falls beyond your famous <5 line directive > > (don't recall the exact phrase you use ;), so we discussed to co-op on > > making an experimental patch to the codebase implementing an inline socket feature. > > I figure this is the route to go for now. If it turns out and and turns out well, > > we might even be able to get you to reconsider. > > Pls benchmark before talking. A thin client written in C is so fast that > its execution time literally vanish compared to the execution time of your > filter. I had something like this that I was using to load an SMTP server > for benchmark testing and it was capable of 600000 runs per hour on a > PIII 900MHz. > I've benchmarked the Scope client to the point of 10000 rounds, and I can tell from this that HDD and CPU makes impact on performance. Without anything to compare the benchmark with, it doesn't tell anything though whether this or that approach is better. /thomas. > > > - Davide > - To unsubscribe from this list: send the line "unsubscribe xmail" in the body of a message to [EMAIL PROTECTED] For general help: send the line "help" in the body of a message to [EMAIL PROTECTED]
