Re: [cobalt-developers] Concurrent Users

["Steve Werby" <steve-lists@xxxxxxxxxxxx>]

"Christian Wieczorek" <wieczorek@xxxxxxxxxx> wrote:
> You served websites to several thousand concurrent users ? Are you sure?
> On ONE machine?

Yes, yes and yes again.

> I think that can't be true. The max. concurrent user value
> apache could handle without a recompile is 512 if i remember right. Even
> WITH a recompile putting the value over 1024 is REALLY hard for most
> webservers and so for a standard RAQ4.

Well, you and I aren't talking about the same thing.  We're both right.  I
wasn't talking about the number of simultaneous client connections to
Apache.  I was talking about the number of simultaneous users visiting a
site.  In a real world situation, some users will be in an active state
(retrieving a page), while others will be in an idle state (viewing a page
that has already been retrieved).  Unless I misinterpreted Jörg's original
post, he was trying to get a feel for whether a RaQ 4 with 1/2 GB of RAM
could accomodate 100 or 200 simultaneous web visitors, not 100 or 200
simultaneous Apache processes.

> Not only CPU time is a point but also there are the Processlist, MEM
usage,
> I/O ... and so on.

Agreed.

> Serving serveral thousand user ... even several HUNDRED users on one
machine
> is a dream wich cant be realised with a RAQ4.
> Building high traffic webservers is no fun and you wont get one by putting
a
> RAQ out of the box. Pure system tweaking!

You're definitely right.  Of course, there are things that can be done to
tweak Apache, limit services running on the box and optimize the code that
can go a long ways.  In any case, I re-read the post I wrote which you
replied to and I want to clarify that I said 200 people can be visiting a
website at the same time on a RaQ 4 with 1/2 GB of RAM (I made no other
assumptions about the usage of the server) and that I said I *expect* one
can handle thousands of concurrent users on said machine (of course, that
depends on a lot of factors).  The last part is based on my own recent
experience with an application I had running on a client's RaQ 4 that was
being accessed by over 1,000 users at the same time (I don't remember the
exact total).  I was able to determine this because I had tracking in place
for analysis of logins and page accesses and each user had a unique id.
About 25% of the page views were of pages written in PHP pulling data from a
well-indexed MySQL db, 20% were static light gateway pages and the remaining
55% were static and heavy on content.  In a nutshell, over the 5 minute
slice of time I analyzed, excluding users who accessed 3 or fewer page
(which excluded ~15% of the users, leaving  ~ 1,300 I believe), the
remaining users accessed an average of about 21 pages (1 per 14 seconds on
average).  I also tracked simultaneous client connections to Apache and
server load, but do not recall what the average and peaks were.  In any
case, the server performed well.  If I have the opportunity to do some more
analysis in the near future I'll be sure to share my results with the list.
Anyone else have a similar analysis they want to share?

--
Steve Werby
President, Befriend Internet Services LLC
http://www.befriend.com/