On Mon, Apr 29, 2013 at 9:54 AM, André Warnier <a...@ice-sa.com> wrote:
> > - under Unix/Linux, there is a command "tail -f <filename>", which > continuously watches for any lines added to a file and displays them. It > doesn't seem to be very intensive in terms of resources used. > - and on the other hand, you probably have a session expiration timeout. > So you could in theory say that you note the start of a session, and then > update this each time there is a new access to that same session. And then > periodically, you go through your table and for each session which you > haven't seen since some time >= the session timeout, you consider it > expired. > > At which time of course I don't know if this is any simpler than the > solution which you are exploring right now. > ;-) > > With all respect, I have to say that looks like a lot of I/O right there and a huge hit in performance and I would assume this would hit memory in a bad way as well, but please correct/enlighten me. Also, I'm following this thread, because I have filter in place for similar reason, keeping track of new sessions created, expiring, etc..., and filter meets my need 100% and don't see why one would want to 'avoid' using a filter. Yes, I was looking at the performance of 'filter' and all that i'm doing in 'filter' on 'every request', but in retrospect, I don't see my filter implementation being much of an hinderance in performance. I have to take another look though and confirm what I'm saying here.