Thanks for the info. I appreciate all of the extra stuff on those patches. I
guess I was most curious about those patches, because if they do work well,
perhaps the concepts they convey could be employed? I'm all for getting a
10x improvement over 1.3 via SSL if possible by using Apache 2.0 on an SMP
machine with a threaded MPM. My earlier tests with 2.0a3 showed, just on
static stuff, that I could easily get a 4x to 6x increase in throughput
using the earlier threaded MPM. It was still a bit broken, but was waaaay
faster than 1.3.x. One thing that has really got the web-business running
scared is the harsh amount of load that persistent encrypted streams seems
to put on a server. I know that with apache 1.3 serving up static stuff via
https, I can get 200 Concurrent users/s pulling the same stuff. I'm
expecting that number to increase dramatically with 2.0 and threading + good
SMP support. I'm predicting that I can get a minimum of 300+/s concurrent
users via encrypted stream with 2.0 + threaded + SMP. The point I'm trying
to make, is as processors get faster and wider, people won't have to be
lured into buying SSL Accelerators for the same price as a server or more,
if there isn't a good need due to better performing software + faster
processors. If you were running perhaps a search engine via encrypted
streams, then I could possibly see the need, even with threading.
Anyway..off my soap box :p
--
Austin Gonyou
Systems Architect, CCNA
Coremetrics, Inc.
Phone: 512-796-9023
email: [EMAIL PROTECTED]
> -----Original Message-----
> From: Ian Holsman [mailto:[EMAIL PROTECTED]]
> Sent: Friday, July 06, 2001 12:00 PM
> To: [EMAIL PROTECTED]
> Subject: RE: Mike Abbot's Patches
>
>
> On 06 Jul 2001 11:11:39 -0500, Gonyou, Austin wrote:
> > Ok..I was just curious cause I was on SGI's site last night
> doing some
> > performance related research..and I saw that those patches
> were still up and
> > remembered some discussion on that some time ago. Didn't
> remember what the
> > final outcome was though. Thanks for the feedback!
> >
>
> The patches on SGI's site are form Alpha-6 of apache2.0, and
> the current
> tree is Vastly different than that now.
> major things I noticed was that things live in different
> places, and the
> patches are pre-filters and use IOL's....
> so It won't be that easy integrating them in now.
> ..Ian
> > --
> > Austin Gonyou
> > Systems Architect, CCNA
> > Coremetrics, Inc.
> > Phone: 512-796-9023
> > email: [EMAIL PROTECTED]
> >
> > > -----Original Message-----
> > > From: Bill Stoddard [mailto:[EMAIL PROTECTED]]
> > > Sent: Thursday, July 05, 2001 6:07 PM
> > > To: [EMAIL PROTECTED]
> > > Subject: Re: Mike Abbot's Patches
> > >
> > >
> > > I worked a few of the simple ones into Apache 2.0. Also
> > > added a 'quick_handler' hook to
> > > facilitate adding the quick cache to Apache 2.0. At this
> > > point in Apache 2.0 development,
> > > you should not expect much in the way of performance. The
> > > threaded MPM scales much better
> > > than Apache 1.3 (uses much less memory per connected client)
> > > but the straight through
> > > instruction path is a bit worse than Apache 1.3 right now.
> > > We'll get that fixed though.
> > >
> > > Bill
> > > ----- Original Message -----
> > > From: "Gonyou, Austin" <[EMAIL PROTECTED]>
> > > To: <[EMAIL PROTECTED]>
> > > Sent: Thursday, July 05, 2001 6:30 PM
> > > Subject: Mike Abbot's Patches
> > >
> > >
> > > > Were any of those patches ever put into 2.0 MPM or 1.3.x?
> > > Just curious,
> > > > don't remember which ones/if any, were ever accepted.
> > > >
> > > > --
> > > > Austin Gonyou
> > > > Systems Architect, CCNA
> > > > Coremetrics, Inc.
> > > > Phone: 512-796-9023
> > > > email: [EMAIL PROTECTED]
> > > >
> > >
> --
> Ian Holsman
> Performance Measurement & Analysis
> CNET Networks - 415 364-8608
>
