Hi, I'm using xcache with a quad (2 dual core) server under Debian. Doc says that count or var_count should be equal to the number of procs. I tried this setup with 96M but php#0 was quickly filled up then started to generate OOMs while php#3 was barely used. I reduced the caches number to 2 and still php#0 is 50% more used than php#1. With this setup I was able to decrease the cache size to 64M without OOMs. I have the same issue but less annoying on the var side. What is the purpose of having many caches? Can I expect some performance increase on the long run or when the load will increase? Why the distribution is not even on each cache? Thank you
on 01.04.2008 03:43
on 01.04.2008 09:30
> I have the same issue but less annoying on the var side. > > What is the purpose of having many caches? for concurrent read and reduce locking in smp box it was left not benchmarked for user to figure it out > > Can I expect some performance increase on the long run or when the load > will increase? > > Why the distribution is not even on each cache? depends on the hashing function, but hashing function do not know the data size. say if you have 10 items and are hashed into 2 caches with 5 items in each, but one of the item is 10MBytes while others are 1KBytes... > > Thank you
on 01.04.2008 09:45
Thanks for the answer Overall the product is very good and decrease significantly the response time.

