Sunday, June 7, 2009

Site optimization

Already planned (some improvements already done) and also in reflection to recent heavy traffic on the site we did some site optimization:
- cache setup
- etags setup
- CDN setup (temporal)
- CSS and Javascript compression
- Pictures compression optimization
- CSS sprites
- MySQL queries and indexes
- conditional layout display

Having all above improvements implemented we were lucky to reduce size of site code by half, loading speed for several seconds and responsiveness of database queries number of times.

The site should be now enough optimized to process traffic of 6 thousand voting users per hour.

Saturday, June 6, 2009

Users attack poll "Creationism vs. Evolution"

Professor at the University of Minnesota, PZ Mayers posted on June 4, 2009 link to ElectionsMeter poll "Creationism vs. Evolution" on Pharyngula (Scienceblogs). He very likely wanted his readers to have a little fun while voting in this particular “pointless” poll. Readers then encouraged themselves to beat the poll to give lesson to all creationists. They came and hit the site with sudden high traffic so the site was overloaded. Within 24 hours 10 thousand users hit the poll (pike 3000 users per hour). To be honest we failed in this test of site performance. The site was very slow responding and even unavailable to some users so we decided immediately to review site code for better performance.

The same poll link appeared also on Bannination site. Where someone posted discussion about Evolution “Proof that Creationists are really juvenile. Do you love Jesus or Darwin?“ and linked the poll.

It was pretty good exercise for us. For the first time we could experience really huge bunch of users using the site, voting and commenting. The only thing we could finally regret after such attack of users (I am sorry to say, fascist of unified opinion) the result of the poll is not very representative. Wow! This is part of democracy. The poll was not abused in any way but simply and perfectly reflects will of the users. A problem however exists with type of users (users of unified opinion). The poll simply and drastically misses users of the other opinion. We as the site providers cannot be asked to adjust the poll result according to any reason, we can only encourage rest of the population to express its vote. The big question is only are there any Creationist, do they know about the poll, do they have interest to join the poll, does the general population ever have interest to vote in poll, if then what the general population be voting for, are only fascists interested in voting?

Will further users balance the result?