im not a noobteck7 wrote:Although yes the tri board has a large amount of kills and EDK doesnt scale well past a certain point, you should nevertheless be able to reach acceptable load times (10s<) per request per visitor given that you:LaMaH wrote:what?..rosander wrote:First off for 98% of the users here the greatest benefit can be found in mysql 5+, innodb tables, and adding helpful index's. 1 Gb of ram with a large cache for innodb tables really makes all the difference and with todays memory prices there is no excuse for not having at least that much installed.
Memcache is helpful to the kb but you do need to weigh the extra install work and upkeep of another program vs using built in cache of mysql. Just to beat a dead horse. For MOST users a well tuned db and cached server will resolve any load issues. Now if you are hosting the kb in a cluster, get memcache. I doubt however many users are. As much as we like to be cool, the kb's just don't draw that much load.
try running an alliance killboard, and then say that a kb doesn't draw that much load
Even WITH Memcache my quadcore server is struggling to keep the Triumvirate board up
1) pay close attention to the optimization details in this thread, in particular the indexing/table format changes
2) use innodb AND MAKE SURE innodb is given a good bit of memory to work with
3) ensure the mysql config is appropriate for both the hardware you are using AND the request load upon the server
4) make sure that web content and databases if possible are being served from different disks (i/o becomes an issue on very active and larger killboards)
5) memcached is nice but careful with how much memory you give it, it will eat whatever is thrown at it often with no benefit to performance past a certain point, you may also want to try PHP APC Cache (http://pecl.php.net/package/apc) along side memcached
5) turn off atime tracking on frequently used file systems (i.e: / , /var etc..), reduces i/o requests (set noatime flag in fstab)
6) running regularly scheduled mysql optimize operations to clear out overhead data is always a good idea , such as weekly (usr/bin/mysqlcheck --optimize --all-databases)
7) make sure the server is not swapping memory to disk (free -m , make sure little to no swap in use), if it is refine the memcached/mysql settings
just my 2cents
me need sleep
Ive done all the optimizations in the thread, innodb has lots of memory, apache is tweaked, mysql is tweaked, memcache is fixed, im using xcache but its the same, weekly optimizations are being done, server is not swapping to disk etc etc etc..
So like, yeah