Well ok, after the election I figured that I could shift the site to some slower storage with a whole lot of redundancy in a RAID6 array of slower drives. We’d have a lot fewer readers than our peak for about six months after post election, right?
Too many of you have been moaning that the site has been running slowly.
Well you are probably aware of what I am like by now. I like numbers and realistic triggers as prompts for action. When I shifted the site to this configuration, I set a trigger.
It said that when we passed 400 thousand page views for a 30 day period I’d look at efficiency improvements. The bottleneck at present is disk access, mainly when the backups run every hour and the search engine updates every 13 minutes. But that spreads out into the post edits and all those comments you leave. Not to mention people simply reading the site. I shove gigabytes of RAM as file and page and object memcaches, as well as great caches on the database, but the system still spends a lot of time hitting those disks.
So when I just looked at this (it wasn’t that way yesterday damnit) ….
Double damnit. More than 400k page views for 30 days and still climbing. That means late night work
Just ignore the Unique Visits, it is a rather meaningless figure because it appears to depend more on the tool used to measure it than anything real. Google Analytics thinks with the same parameters that it is about 120k. Sitemeter usually thinks it is higher. Whereas all of the tracking tools agree within a few thousand what the page views are after they discard all of the bots. It is a much more useful measure.
So why was that value of 400k page views the significiant trigger? Well if you look at the monthly figures for last year (I put the values in this one to make it clearer)…
Damn you readers. I bloody well know it is you because we don’t use the steroid site enhancers of some other sites. I can see you buggers causing me late night work……
…. I’m going to keep an eyeball on you just like philu*….
* Except without the side effects