GIGAZINE was interviewed on "Web expert know-how Vol.4"

"Targeting the business purpose web site operator and others as the main target"Web personnel site know-how"There is a feature called" Waza of Access Improvement 99 ", in which GIGAZINE was interviewed. By "GIGAZINE flow access improvement point", it is briefly summarized.

So, for this time, I will disclose how much improvement resulted as a result of heavy access due to flood of access and moving the server.
By the way, the interview article of GIGAZINE is posted on page 49.


So, I think that you can tell how much improvement you made by moving to a new server in the daytime, in the evening, etc. by the sensible speed, but let's see how much improvement has been made as a numerical value. The graph is "MuninIt is what generated it.

First, CPU load, so-called load average. It has become very terrible. It is too heavy.



As a result of switching to a new server, this has been improved. The maximum load was reduced by one digit.


You can see how improvement is improving by summarizing the CPU usage rate and so on. First of all is the old server. I already exhausted resources in every sense.


And, a new server. It's like this because you have two dual-core Xeons.


Thanks to the old server, swapping also occurred and it caught a vicious circle and it was such a tremendous load. Especially September before customizing the settings of the server is terrible.


Also, since the MySQL version has changed from 3 to 4 on the new server, we turned on query cache. Because it hits quite a bit, the load of MySQL significantly lowers.


By the way, if Apache, MySQL, PHP, and the parts that have modified the script itself are left as the default setting, the load will increase from 10 times to 20 times this load. If we accumulate such know-how, we can distribute the remaining budget up to around cold standby or disaster recovery without increasing the number of servers when wastefully load balancing with multiple machines.