I have followed with interest a number of comments about the reliability of Cloud Computing, based on the actual, or perceived, problems with service from Google and Amazon. Of course there are some real incidents that have caused outrage, (my definition of the rage one experiences from a system ‘outage’), but are failures in ‘best endeavours’ mail services such as Gmail really the basis for considering the risks and rewards for Cloud Computing? May be not, but more serious was the outage of Amazon S3 storage service Sunday July 20th. This was the second outage from a respected leader who is selling cloud capacity as a service.
Amazon is well respected for its technology capabilities so this was a shock especially as a month or so before they had apparently been blaming problems on their ‘complicated systems’, but then for Google to have had a whole series of issues in the mid august period that affected some of their other services too. Taken as a whole against total system down time the figures aren’t that bad, but it does rather attract your attention when thinking seriously about the topic and its role in Enterprise services over the coming couple of years.

When you think about it carefully your realise that there are a lot of unknowns in what happens as there are an increasing number of variables that you don’t see in a simple SaaS, or Enterprise service. On this basis it makes you wonder how Microsoft will get on with their online small business service. My guess is that this is built around a more controlled environment, but as a full enterprise service on which those small businesses who sign up for it will rely I hope this is true. This looks like one place to watch progress carefully, along with the eye I have been keeping on SalesForce.com.
The real moves I want to keep an eye on are the tie up between the big boys Intel, HP, Yahoo and some good academic computing centres with the aim of building a really large scale test bed for testing and experimenting with Cloud Computing. The challenge with Cloud Computing is that the ‘business case’ is compelling, and I don’t just mean costs, I mean as a method to deliver the kind of Web 2.0, Enterprise 2.0, Business Technology solutions that offer some real new business value.
Many of these solutions will eventually require ‘Orchestration’ between services to take place in a Cloud Computing environment. Put another way the environment will be expected to support a lot of previously untried combinations of activities on the basis that all of them subscribe to the same common principles. Right now, you couldn’t say that we have those principles well enough defined for this to have a sporting chance. That’s what interests me about this project and its trials.
After all Virtuality looked a pretty simple trick, so who would have guessed that when it moved into the real mass market even a leader like VMware would get caught out. If you haven’t heard they have a clock drift problem that means time stamping within an application can drift off. Pretty critical in trading applications in the financial sector! It’s the small things that matter in all of these approaches!!