Am Montag, den 05.03.2012, 11:47 -0800 schrieb Luke Scott: > So for the most part there really isn't much going on after a request has > been made. A visitor requests a page, sits on it for 3-5 minutes, submits > the form, and then leaves. But the more users we have and the more pages > they have the higher chance of requests happening in parallel. Right now it > doesn't, unless a user gets a huge surge in traffic (Digg effect).
Do you think the following model would work? - have a limit for the number of workers: 100 - one worker or zero per user (not visitor!) - when a worker is needed but none is active for the polls/forms/... creator, kill the one with the highest idle time and spin up a new one (EXCEPTION: you'll have to temporarily go over the limit if all processes are busy) So, when you get a huge wave of requests for a bunch of forms, the sandbox processes for them are running all the time. Does that sound like a good idea?
signature.asc
Description: This is a digitally signed message part
