Amazon Web Services issues highlight the risk of putting all your eggs in the cloud basket

Three lessons for business continuity managers which arise from the recent downtime incident which affected Amazon Web Services.

Amazon Web Services experienced almost an hour of downtime on Sunday due to
‘connectivity issues’ in its North Virginia data center. The hosting company is one of the largest providers of public cloud infrastructure in the world. According to a report by Deepfield Networks, websites hosted by Amazon Web Services are visited daily by a third of all Internet users; and the cloud infrastructure is also used by companies to host enterprise workloads.

The above makes Amazon Web Services a single point of failure for many organizations: and the weekend’s outage shows that despite a huge investment in resiliency the company’s network is not infallible.

What recompense will Amazon customers be given? The service level agreement only guarantees 99.95 percent uptime; which sounds good but actually equates to 4.38 hours downtime per year. Downtime beyond the 99.95 percent level will result in a service credit of either 10% or 30% depending on the actual length of downtime experienced by each client. However to receive the 30% credit, downtime must drop to 99.0 percent, which is a highly unlikely 7.2 hours of downtime in a month.

Such levels of recompense are unlikely to scratch the surface of the associated costs of downtime to enterprise users.

Three lessons for business continuity managers seem obvious from the above:

The public cloud is not yet reliable enough for mission critical processes; and these include any processes which will be relied upon for business continuity or disaster recovery purposes. The perfect storm of a disaster affecting an organization and coinciding with a cloud outage preventing access to business continuity resources would be unlikely, but at current levels of public cloud availability levels, it is far from impossible.
Day-to-day production computing operations which are carried out using cloud infrastructure, or which rely on cloud-based software as a service, must be capable of being carried on off-line should the cloud be unavailable. And the cut-over needs to be relatively seamless.
The details of the service level agreement which exists with cloud providers needs to be understood; as well as their actual service history. Different cloud providers offer different levels of uptime and also have varying standards of customer services following a downtime incident. Historically, how much downtime has the cloud provider experienced? What reviews did customers give about customer service during service outages? How quickly were services restored during an average outage? The answers to these questions will give a better idea of the true levels of downtime that can be expected.

Make a comment

•Date: 27th August 2013 • World •Type: Article • Topic: Cloud computing

Sign up for Continuity Briefing, our weekly roundup of business continuity news. For news as it happens, subscribe to Continuity Central on Twitter.

How to advertise on Continuity Central.

To submit news stories to Continuity Central, e-mail the editor.

Want an RSS newsfeed for your website? Click here