Rackspace UK needed a thorough testing, so as a full blooded British person, I have tested it for you. I love to test everything and anything. I even tested the theory that man can walk on water, and I did! (the pond was frozen).
The public cloud is the possibility to have access to unlimited computing power, and you only need pay for pay for consumption. Considering that it’s use is unlimited (across home, business, or even the country), you pay only for what you use. Like, have you ever tried to convert a file with your installed file converter and it decides to shut down your media player because it needs the CPU power? Well a thing like Rackspace UK would make all the effort so that you could continue watching your DVD of famous world cheeses.
There are mainly three types of cloud services. There is the laaS Server, such as Rackspace, AWS, Joyent, etc. There is the PaaS, which is storage based on a particular language, so it would be Google Engine with GWT, Azure.Net with Heroku leRuby, etc. The final type of cloud service is the SaaS application, such as Google Apps, BaseCamp, etc.
The build is simple, i.e. a web server (Nginx), PHP servers (FPM) powered by a load balancer (HAProxy), and a database (MySQL). The shop was built on the open source Prestashop solution.
If a users most important quality is the number of visitors, then obviously the PHP and the database that will suffer the most (functionality wise, aka, the probability of overload/crashing). I have defined the two scenarios as firstly servers PHP / FPM in the Cloud and secondly servers PHP / FPM and the MySQL database in the cloud. Using my benchmark tools, I simulated visits on the site, ranging from 1 to 1000 concurrent users (1 visitor is considered a psychopath who refreshes one page per second). It’s what losers used to do to increase web traffic numbers before the tracking programs got smart.
The cloud choice was rather simple; they were Amazon Web Services (AWS) and Rackspace UK. After two weeks of intense testing and optimisation I discovered that, the CPU number in Rackspace is limited to 4 (as per my testing). Was this compensating for my needs or was that the upper limit? I’m not sure.
It is sometimes possible to stop a server at AWS (and not be charged), this is not the case with Rackspace. This is not a big loss unless you happen to be fairly indecisive. The MySQL performance of the RDS hosting database is below what some people have on their own machines, however the hardware is weaker at home. It is a little like complaining that your car stereo is not a fantastic as the one you have at home, but in the same way your house cannot move as fast (or at all, unless you have a camper van, but you don’t count towards my analogy).