Managing traffic spikes during the 2018 World Cup with the OVH Public Cloud

As the FIFA World Cup kicks off in 2018, websites dedicated to the national sport will experience their highest traffic spikes. Nicolas Capeyron created Sysadmin Badass in 2009. His job is to administer web-oriented servers and manage his customers’ infrastructure architectures: e-commerce websites, tourism websites and SaaS-based software publishers. For one of this customers, French internet media website sofoot.com, he decided to use OVH’s Public Cloud solution. Here, he explains the reasons behind his decision.

Nicolas Capeyron: I am the administrator of www.sofoot.com, a news website dedicated to football, which has been hosted by OVH for many years. The website’s visitor count is growing and, during the 2014 World Cup in Brazil and the recent Champions League Final, we needed more capacity than the infrastructure could provide. We usually have around 100,000 to 150,000 different visitors per day. This number can double, or even treble, during big events like the FIFA World Cup. At these points, the number of requests per second can be up to 20 times higher than average. This is because SoFoot live updates and its match analyses are really popular, and people comment on them a lot. The capacity to withstand traffic spikes is, therefore, essential to our business.

sofoot_logo_sysadmin_badass

So did you choose the Public Cloud in order to handle these traffic spikes?

N. C.: So far, during the World Cup and Champions League, we migrated the website to a more powerful front-end, then we implemented some optimisations. The procedure was fairly simple and inexpensive, but it had certain limitations. For example, we couldn’t guarantee that the traffic would stay within the server’s technical limits. The website didn’t have high availability or load balancing, and with limited development time available before the World Cup, we could only do the bare minimum in terms of technical adaptations. However, we had to get the most flexible and elastic infrastructure we could find, so that we could absorb sharp spikes in traffic. So I took a look at the various cloud solutions on the market. I discovered OVH’s Public Cloud solution during the tennis World Tour in Lyon, tested the service, and thought: bingo!

How have you integrated the Public Cloud into your architecture?

N. C.: We have built an architecture that is designed to offer both high availability and cost savings. The IP Load Balancing service shares the requests between two dedicated servers equipped with HAProxy. In turn, they distribute the traffic to three other physical dedicated servers, to which five Public Cloud virtual machines of varying weights are attached. We have added the availability of the IP Load Balancing solutions and the OVH network to HAProxy’s powerful configuration, in order to address all the front-ends we have at OVH. This way, we are able to serve all of the website’s 60 GB of static data very quickly.

So does the Public Cloud meet SoFoot’s requirements?

N. C.: Yes, we are really satisfied. Network and storage performance rates are impressive, and these are crucial factors for SoFoot. Each of our front-end servers can now support both the website and a copy of all the static data. Obviously, it also absorbs the additional loads during traffic spikes. Moreover, the choice of a standard like OpenStack allows me to use regular DevOps tools like Vagrant or Puppet. In short, we get the benefit of excellent performance, an innovative cloud solution and the best value for money on the market. If we had tried to get an equivalent service from one of OVH’s competitors, we would have ended up with a significantly more expensive solution, or even had to overhaul almost everything we had put in place to make the website work.