Re: Estimating number of users with number of IPs

From: Hyunchul Kim <>
Date: Fri, 12 Dec 1997 00:48:16 +0900 (KST)

> On Thu, 11 Dec 1997, Hyunchul Kim wrote:
> > Greetings!!
> > i'd like to ask a question about the relationship between
> > the number of users and IPs.
> > maybe it could be calculated as .... (IPs * X), i think.
> > does anybody know the reasonable value of X or other formula, if any?
> "X" depends on your user population, I guess.
  of course it is.
  i mean, "rough(not exact) approximation or order of X"
  which might have been used in many internet domain/population surveys
  for calculating/estimating internet user population with the number of hosts
  they already had acquired.

  Maybe....5 - 10? or more?

  hyunchul kim


For example, if your Squid
> serves PC clusters at some university, X will be very large (30-40?). If the
> cache is for research workstations in the CS department, X will be close to 1.
> If you have a mixture, you can calculate X if you know the proportion of the
> components.
> For a cluster-like scenario, I believe you can validate your calculations by
> using a fairly simple script that counts the number of "startup pages"
> requested from Squid during the day. It is probably feasible to guess what
> the startup page is for every regular client by analizing access logs for a
> week or so. It will be a page that "often" starts a sequence of requests from
> a particular client after a "long" idle period. Counting just the number of
> idle periods might work as well provided your clients do not "sleep" at the
> terminal with the browser running.
> Also, in many cluster-like environments, you know a priory what the startup
> page is (e.g. a University Home Page).
> Alex.
> P.S. I assume you are talking about leaf proxies.
Received on Thu Dec 11 1997 - 08:08:26 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:37:55 MST