Hi,
in one of our implementation we deliver like 600/700GB of traffic every month, for 1.5 Million hits/day and 300K page views/day.
It's a cluster of two 2-way servers with an additional server doing reverse proxy with squid and with postgres installed on a different 2-way server.
We use OACS5, AOLserver 4.0.10, tcl 8.4.11.
The system is performing really well, but there's an issue that we are still trying to address. Aolserve+OpenACS seems to consume on the average 15MB per connection/thread and this really seems a huge amount of memory, especially because it grows linearly with threads (200 threads ~= 3GB). It seems that almost nothing get shared among threads.
Anybody of you could share with the forum the average amount of memory consumed by your implementations and/or any idea to reduce/track memory usage?
Thank you, ciao,
Giampiero
