Yesterday morning, millions of websites were off-line as fire raged through a data centre in Strasbourg. OVH is the fourth largest web-hosting provider in Europe (after AWS, Microsoft Azure and Google Cloud), and operates four data centres on the Strasbourg site. One 500m2 data centre was destroyed and another was damaged in the blaze; the other two data centres were taken off-line but appear not to have been damaged. Incredibly, OVH aim to have three of the data centres running again by the end of next week. In the meantime though, high-profile customers, including the French Government, the European Space Agency, and the Pompidou Centre; have been severely impacted.
Whilst such dramatic incidents are extremely unusual, disruption to data centres is surprisingly common. For example we have blogged in the past about:
- Failure of both power supplies to a Telecity data centre in London in November 2015;
- Failure of a BA data centre due to “human error” in May 2017; and
- Failure of both data connections to a Hosting UK data centre in St Asaph in June 2018.
Research in 2010 found media reports of 32 major outages (average duration 15 hours) over a 30-month period. Based on this level of publicly-reported incidents, the authors estimated an annual 2.5% chance of a major outage for a typical commercial data centre.
As we often say: you can outsource a process, but you can’t outsource the risk. Whilst major providers of hosting services are extremely reliable (probably much better that you can achieve in-house), they are not infallible; so you must still have plans in place if they fail. As OVH Chief Executive, Octave Klaba wryly observed in a tweet to customers yesterday, “We recommend to activate your Disaster Recovery Plan.”