On Fri, Jun 13, 2008 at 3:59 PM, [EMAIL PROTECTED]
<[EMAIL PROTECTED]> wrote:
> Are main CPU hog is the Captive Portal, with 50-100+ people trying to login
> at the same time, it can eat up the CPU big time. If i turn captiveportal
> off, are 5501s barely peak over 30% cpu, with it on, I'm seeing 100% spikes
> all the time.
>
Yeah I don't think it's on the hardware sizing page on the website
yet, but in the coming book on pfSense I have included some info on
hardware sizing for captive portal. Here is an excerpt that needs some
touching up still, and will be on the website when finished.
Large and Busy Captive Portal Deployments
Captive portal deployments with thousands of users and/or frequent
simultaneous log on and log off activity will require more CPU
than is required under normal NAT or routing operation. The processing
of user log on and log off events, as well as maintenance of the user
database increase CPU usage to some extent. How much depends on the
total number of users, and most importantly the number of users logging
in simultaneously.
We know of several universities, schools and businesses around
the world that have thousands of captive portal users on a single
server. The deployments we are familiar with use moderately recent
server hardware with dual Xeon 3+ GHz processors, and have plenty of
CPU capacity to spare.
> I can't see me ever having a pipe bigger then 50mb/s or a DS3. So I'm pretty
> sure the box will be able to handle that throughput without a problem.
>
Never say never. :) The hardware you specified is more than adequate
for a few thousand users with a couple hundred frequently logging in
simultaneously and still providing > 100 Mb of throughput, so even if
you do have a bigger pipe eventually you should be able to scale
nicely.
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
