On Mon, 26 Sep 2005, Henrik Steffen wrote:
but one could perhaps easily consider to change the
statistics of "Number of clients accessing cache" into
a number counting the absolute clients since last restart.
For what use?
In an Intranet where the number of clients is limited such number will
start at 0, and over time grow to the number of stations you have on your
network, and each time Squid is restarted the number resets to 0. The
number of stations you have you probably already know reasonably well, and
any number inbetween until the count has stabilised is not very useful for
any purpose.
Another problem is to keep track of this in an efficient manner. To keep
this statistics Squid needs to remember each and every IP address which
has sent even a single request to the HTTP port. If your Squid is
connected to the Internet this can become quite many addresses to keep
track of.. such counter tells you absolutely nothing about the recent
traffic, only how many IP addresses have ever been seen by the proxy. For
a more indepth reasoning why this was changed to the current behaviour see
<url:http://www.squid-cache.org/Versions/v2/2.5/bugs/#squid-2.5.STABLE6-client_db_gc>
This would go together perfectly with the
"Number of HTTP requests received", which is
an absolute number as far as I know.
It is, and as such is practically useless unless you sample it and look at
the derivate function which is highly interesting (gives you the number of
requests per the sample period).
The derivate of "Number of clients accessing cache" as an absolute counter
since last restart on the other hand has no or very little practical
value. Immediately after restart it will be quite high, but will quickly
decline and after about 24 hours it will become 0 in most networks.
The derivate of the current "Number of clients accessing cache" has some
meaning. If you see it increasing you know you have new clients who
haven't accessed the proxy in a while
What use is a statistic, where nobody knows what's actually counted?
As I said the number gives a reasonable estimate of the number of
users/clients (per IP) of your cache today. I can provide you the exact
algorithm if you like (it's mainly in cliendbGC), but it probably will not
make you much wiser.
In my eyes the absolute "Number of clients accessing cache" since restart
is even more confusing.
Regards
Henrik
