As many people will have noticed, the Top 10 read diaries is now in a random order and the number of hits no longer shown. The reason for this is quite simple and that is that the listing is there to give new people (and perhaps not so new people) an easy starting point to read the popular diaries. It wasn’t intended as a competition that one should win by any means fair or foul. And thats not meant to be suggesting anything at all about anyone. So now it serves its intended purpose and the element of competition is removed.
Some people like it. Some don’t. To those that don’t, well, we’re sorry you don’t but please don’t expect it to go back the way it was. As we’ve oft been told, it’s our site and we let you use it 🙂
On other issues we finally managed to get PHP working with MySQL and the PAM modules. I shall start a development diary for a little project of mine shortly, but anyway, we’ve been looking at moving DearDiary away from perl and into PHP4. Except it wouldn’t work with the existing setup which would mean another MAJOR overhaul to the system. Which I pretty much veto’d at the moment on time grounds 🙂 But if the PHP can be migrated in gradually (which it now can) then I’m all for it and happy to help develop. So that was cool.
As part of that investigation we upgraded to the latest version of Apache and nearly died as a result… The process size for each httpd process had multiplied by three from 20M to 64M or so. Not bad huh? Which meant there was room for one more httpd process before the server started swapping to disk and slowing down. I resolved to find a way of reducing it back down to the 20M process size it was…
I did better than that, its now a whole 4M each. The server has so much memory now that its bored most of the time. Which is how it should be. There shouldn’t be any more problems with it running out of memory or swap space for a while… For those that care we now have enough room on the server for around 250 running httpd processes (since each process actually uses only 1M of memory exclusively for itself, the other 3M it shares with all the others). Given that on average it takes a LOT less than a second to deliver a page, but lets say its two seconds to account for increased load, this means we could serve approximately 100 requests per second. Each page requires around 4 requests, so thats essentially 25 pages per second we should be able to deliver. Or 1500 per minute… Or 90000 per hour. Or around 2 million pages per day. If we’re serving 2 million pages per day I won’t mind buying another server to help spread the load 🙂
Come to think of it, if we’re serving 2 million pages a day you can probably assume you’ll need to page me to my phone by the beach. Somewhere warm, white sanded and with palm tree’s.