Not convinced to move to the Cloud (yet)

2011-06-20T12:03:00.000-07:00

Cloud computing is currently very topical and could have a huge impact on the way we structure our hosting business. We moved from shared hosting back in 2003, originally co-locating a single server running Windows Server 2003, and soon moved to a multi-server suite with separate servers dedicated to email and (most SQL Server) database web applications. We no longer buy our own servers but rent what we need as when we need it. Redundancy comes solely in the form of mirrored hard drives. If a server fails, we'd need to rebuild a replacement from scratch. Because these are rented we can get a new one up and running in under an hour - a far better scenario than when we would have to have a new one shipped in to the data centre - but still not a situation we relish. We did lose a server once, our main SQL Server machine and it took 48 hours to fully recover. (Now they sit in a sorry pile in the corner of the office.)

Having reliable backups is key of course and this is one area where we have embraced cloud services. We are just starting to resell a very efficient cloud backup service and also use another cloud backup for backing up our servers. These have the advantage of offering remote storage at reasonable cost and we can be sure that redundancy is built in. At the end of the day, if these services were unavailable for a day, it would not be a critical issue - unless it coincided with at a failure at our end.

With other services though, it's a different matter. If we were relying on a cloud service for web and/or database hosting, a failure is going to have an immediate impact on our clients - and there would nothing we could do except wait for the service provider to sort out the problem. A classic example of this is the failure of
Amazon's EC2 service in April (see cnet News) which brought down some major sites like Reddit, Foursquare and Quora.

The lack of control would be a concern and not just in terms of QOS. Currently our hosting costs are fixed. If we have a server that's under-utilised we're probably paying over the odds for storage and bandwidth but it's better than an open book. Because we are charging our clients at a fixed cost we can't move to a usage-based model without risking our profits being wiped out by unexpected storage and traffic peaks. And the costs just don't compare to a simple dedicated server. Pricing Microsoft Azure for hosting 50 SQL Server-based web apps comes up with a base cost more than twice what we are used to without even factoring in the bandwidth.

So this week we'll be ordering another dedicated server to relieve the load on a two year old server and enjoying a cloud-free weekend 8-)

ReZolve Developments

Not convinced to move to the Cloud (yet)