Interview with Dmitry Sotnikov, COO of Jelastic

By Dmitry Sotnikov (Profile)
Wednesday, March 20th 2013

VSM: What do you make of Windows Azure facing yet another outage recently?

DS: I wouldn't blame Microsoft too much here. Obviously, outages are no fun, and all businesses relying on the Azure cloud were affected, but I don't think that from an outages perspective Microsoft is much worse than, say, Amazon.

VSM: That's right, 2012 was a rough year for Amazon.

DS: In the last 12 months, Amazon had quite a few high profile outages affecting all their brand customers like Heroku and Netflix.

VSM: Does this mean that "cloud" is just not for applications that need high availability?

DS: The reality is that any hosting from any company, including hosting done internally by your own IT, can and will go down eventually. When you are choosing your hosting platform and are designing your application, you need to take that into account.

VSM: Are you talking about the geographic redundancy and availability zones that Amazon and Azure have?

DS: Only to some extent. If you look at some of the recent outages, they did not actually seem to help even the customers that were using them. A platform from a cloud company is in this case still a single live service and it inevitably has interdependencies between parts of the platform in different geographies that can lead to cascading effects, or produce bottlenecks and points of failure.

VSM: Is there any way that customers can increase their availability?

DS: Sure. Use more than one service provider. Obviously, you can only receive Amazon Web Services from Amazon, and Windows Azure from Microsoft, but look at what Apple famously did for their iCloud. They designed the system to use both clouds, and unlike Netflix, they have not been affected by outages of either of them.

VSM: Is it that easy?

DS: Actually, in this particular case it is not easy at all. AWS and Azure are very dissimilar, so your software development budget needs to be significant to bear the overhead of creating and maintaining the service that spawns such clouds.