wamp server 2.0 on vista 64. I really have no way of replicating this over on Mosso, although I think they would send me an email if i was bringing down part of their cluster.

I am not sure why I am getting this issue as this is outside my realm of expertise and area of interest.

Although I did make a fair share of money on .net throughput by implicitly declaring objects null after they were done being used instead of letting the "runtime engine" and garbage collector come by and sweep it up.

We had this one issue with silverlight and we dropped the cpu usage down 45% by nulling everything after we were done with it.

I don't know if filesystem cache is the issue, but probably not. I know we always put stuff in httpruntime cache which was in server memory and that worked very well, but we had intellegent cache management where if the cache got over 200mb it started deleting stuff that wasn't frequently used. I am not sure that php has that sort of thing, but I am guessing eaccelerator has that functionality...I just haven't dealt with it.

never had such problems.

are you using eAccelerator? It make use a lot of memory because of the cache.

I've been using c5 on a lot of different servers, but lately only on debian
- Apache/2.2.9 (Debian)
- PHP 5.2.6 (cli) (built: May 14 2008 09:46:47)

got plenty of free memory on my servers (usually about 3gb free) but I'm still watching it and it basically never changes.