Re: Estimating Users

From: Alex Rousskov (rousskov@measurement-factory.com)
Date: Thu Jan 25 2001 - 20:46:54 MST


On Thu, 25 Jan 2001, Charles Hapsburg wrote:

> I am using polygraph to "stress test" some squid/linux proxy
> servers and was wondering if there is a good way to take the
> requests/sec and error rates and extrapolate an estimated number
> of supported web clients? So if I have 120req/sec w/ 0% errors,
> how many users proxying through the machine does that translate
> to? (assuming that they are all doing basic web surfing) I
> understand that there are a tremendous number of variables that
> could affect the results, but I am just looking for a rough
> estimate.

My guess that an average "busy" Web surfer visits about 2-3 pages in a
minute. Let's assume that each page has 10 embedded objects, on
average. That gives us about 0.3 - 0.5 requests per second. PolyMix
robots happen to use 0.4 req/sec mean rate.

I have no idea how the number of active users would correspond to the
total number of potential users though. Clearly, not every ISP
subscriber surfs at the same time. The ratio is probably dependent on
the user population (ISP vs. corporate vs. library). Moreover, the
increasing number of various non- and semi-human surfers is not going
to make the traffic patterns easier to grasp.

$0.02,

Alex.



This archive was generated by hypermail 2b29 : Tue Jul 10 2001 - 12:00:17 MDT