(posted at 09:43PM GMT)
I am now down to the last hundred or so sites left to move across to our new shared webserver (see my January post on the subject).
There are currently just over 3,500 sites on the box; of which, just over 75% of them use PHP and over 90% of those use a MySQL database (which is provided via dedicated Gig-E interconnect to a box which just does MySQL).
Wanna see the load averages ?
[admin@fluffle ~]$ uptime
21:47:40 up 60 days, 12:41, 1 user, load average: 0.00, 0.00, 0.00
The box pushes an average of 6Mbit/s at peak times dropping to about 2-3Mbit/s during off-peak hours - yep, it is currently running half-load but still only has a load average of 0.00.
Compare that to a box running Apache/mod_php (not ours but managed on a customers' behalf) with a few hundred sites on it:
[xxxxx@xxxxxxx ~]$ uptime
21:43:41 up 245 days, 6:13, 1 user, load average: 0.54, 0.53, 0.95
... and a similar box with approx. same number of sites running Apache/mod_php/suPHP:
[xxxxx@xxxxxxx ~]$ uptime
21:52:00 up 23 days, 13:07, 1 user, load average: 0.77, 0.72, 0.55
On average, the two Apache boxes combined do less than 30% of the traffic of our shared webserver.
To be completely fair and to ensure that MySQL isn't unfairly blamed for the excessive load averages on these boxes, I'll throw the load averages from our dedicated MySQL box on the fire:
[admin@sentinel ~]$ uptime
21:52:59 up 44 days, 9:46, 1 user, load average: 0.00, 0.00, 0.00
The dedicated interconnect runs at about 10-15Mbit/s sustained at peak rates with it dropping to around 8-9Mbit/s during off peak hours; quite a few phpBB forums on our shared webserver and the search functions do get used quite heavily.
Based on the above data, it certainly looks like Spilsby Internet Solutions is 'technically leading' the way when it comes to high performance/low latency web hosting ;-) |