counting concurrent users...need advice.
Fairlight
fairlite at fairlite.com
Fri Aug 27 18:21:55 PDT 2004
On Fri, Aug 27, 2004 at 09:08:50PM -0400, Bob Stockler may or may not have
proven themselves an utter git by pronouncing:
>
> I thought you were looking for something that would run for
> just a litte while to tell one how many user licenses to buy.
Well, no, not really. :) More a case of, "We already have 'x' licenses,
spread across two or three clustered web servers. Now, at any given point,
what is the maximum count taken on each server, and what -is- that count.
And when do those peak periods occur?"
> wouldn't be too expensive to run, even if the sleep time were
> reduced to 5, 4, 3, 2 or 1 (or maybe even eliminated).
If you eliminate it (you'd need to in order to discover sudden spikes, and
you -still- have the potential for a race condition where you get a spike
and miss it between two consecutive poll/parse runs. It's not as foolproof
as I'd like.
> Bob (who, admitedly, doesn't undestand the problem)
The problem is needing to be able to catch the user count at any given
point of granularity--basically in realtime with -no- granularity is ideal.
Once you have an assurance that you have the most data it's technically
possible to obtain, you can worry about taking the high point per "window"
if you wanted a graph of every quarter hour or whatever, if the "when" part
of it was terribly important. The crucial part is the "how many" portion,
which is where the finest possible granularity is critical--especially when
running a mix of slow and fast jobs in a web-based environment where data
entry doesn't tie up 95% of your running time.
mark->
--
Bring the web-enabling power of OneGate to -your- filePro applications today!
Try the live filePro-based, OneGate-enabled demo at the following URL:
http://www2.onnik.com/~fairlite/flfssindex.html
More information about the Filepro-list
mailing list