We are having 5 web servers and 1 db for our board.
Our board generates about 20+M pageviews/mth with approx 1300 - 2500 users online (as reported by vB on a 900 secs timeout)
Anyone can share with us on how to optimize our setup? Our servers are installed with off-the-box RHEL4. We do not have knowledge on how to custom build kernels or modules.
Apparently that way seems to be the most efficient to keep load down and serve out fast.
Anyone can share where to find such info.
For those who indicated 2 dbs, how is your setup like? How to have 2 dbs in 1 vB installation?
Big-boards rank ca. 250 and going up.
Alexa rank ca. 12,500
2000-3000 users online 24/7
4,7 million posts
135,000 members
One server:
RHEL 5 64-bit
Sphinx search
Litespeed Enterprise Server 3
2 x Intel Xeon-Woodcrest 5148
8GB RAM
2x SAS-SCSI (15k rpm) + SATA + NAS
Server can handle much more than above needs, it's on light load now.
--------------- Added [DATE]1200784410[/DATE] at [TIME]1200784410[/TIME] ---------------
Quote:
Originally Posted by TECK
With 5,000 users, you need 6 servers.
Sorry but that is simply not true.
Quote:
Then your server load will be around 0.5-1.5. You cannot run an efficient board while having the load at 15.
Load is essentially "processes in queue". 1*cores load means all works in about realtime. To interpret any loads higher than that is not that straighforward. What those processes are, are their queuing really slowing things down, how long do the queued processes need to execute etc?
Would you mind share us your config for Litespeed? i have a board with that much of member online but less post and thread. Running on Xeon 5335 with 4GB of ram and 73GB SCSI for both webserver and database.
Big-boards rank ca. 250 and going up.
Alexa rank ca. 12,500
2000-3000 users online 24/7
4,7 million posts
135,000 members
One server:
RHEL 5 64-bit
Sphinx search
Litespeed Enterprise Server 3
2 x Intel Xeon-Woodcrest 5148
8GB RAM
2x SAS-SCSI (15k rpm) + SATA + NAS
Server can handle much more than above needs, it's on light load now.
--------------- Added [DATE]1200784410[/DATE] at [TIME]1200784410[/TIME] ---------------
Sorry but that is simply not true.
Load is essentially "processes in queue". 1*cores load means all works in about realtime. To interpret any loads higher than that is not that straighforward. What those processes are, are their queuing really slowing things down, how long do the queued processes need to execute etc?
I can see this. I can't believe the load difference since we switched to a SAN. I would think NAS would have the same functionality. It really seems to be all about the drives. I'm not about to try it but I bet I could get away with not having seperate front end servers.