If I had a fiver for every time I’d heard a tale of woe about a cloud deployment that had ended up costing more than expected, I’d almost have enough to put a down payment on Luis Suarez.
It’s a plain fact that the gap between the theory and the reality when it comes to cloud is a big one. We’re dealing with a whole new area, one where there really is no guidebook and few explorers who have made the trip. One thing that is clear is that there’s little point in taking what’s happened on-premise as a guide: the differences in technology could be stark – a system built around virtualisation can eat up storage, for example.
But it’s not just the technical differences: cloud, by its very nature is a pay-as-you-go system – that means being diligent about closing down instances.
Users have not had to bother about this in the past and, now, all of a sudden it has become vital. It’s like students leaving the parental home. They go from somewhere lights blaze 24-hours a day, the TV never gets switched off and a teenage room has more electronic gadgetry than the Starship Enterprise, to a shared house where the quarterly electricity bill is pored over with zeal.
This is why cloud services have not delivered what they've promised. There's a human element that can make a mockery of any planning. No matter how carefully a CIO plans for a move to the cloud, something will always crop up to make the projections seem like utter nonsense.
That’s why it was interesting to talk to Hassan Hosseini of Plan For Cloud, a company swallowed up by Right Scale last year. Plan for Cloud offers a tool that enables CIOs to look at their cloud plans and ascertain how much they’re going to spend. Every cloud provider in the book claims to have a similar tool, but the difference with Plan for Cloud is that it covers several vendors – not an easy task as there’s no standard method of charging for this or comparing costs between cloud providers.
Sooner or later, this is an issue that needs to be addressed. It’s still early days as far as cloud is concerned, but there will need to be a standard model for cost comparison. I’m reminded a bit of the early days of mobile telephony when it was near-impossible to weigh up the benefits of one provider over another, thanks to the different accounting and charging methods. Measuring cloud workloads is an issue that I've mentioned before and one that I suspect will crop up again.
Hosseini says that he was used to seeing Excel spreadsheets from companies whose proposed budgets were way out of kilter with what they were actually paying. “A spreadsheet isn’t going to do it; our research showed that moving to cloud could involve 12,000 price pressure points,” he says. “For example, we found a company in London who thought compute was going to be its main cost and allocated little to storage as that would be hardly used. This bit was true – there was a low volume, but it was accessed regularly and the storage costs started mounting up, simply because of the cost of access.”
The other problem that Hosseini found was that users weren’t being monitored sufficiently, with many companies having no controls in place. “We found there was a need for automatic processes to shut down services when they weren’t being used,” he adds.
Cost is just one part of the cloud equation and customers should not get too hung up onthe price tag of their cloud service. “Cost is one aspect. There’s also performance and feature-set to consider," adds Hosseini. "Cost we have nailed down, but we don’t cover the others. If you want to do a real comparison you have to look at other aspects of the service.”
Having a cost-comparison tool is a good start. Indeed, giving users a decent estimate about the cost of a service will be an excellent tool for future cloud deployments. The only question is why has no-one done this before? It's understandable that AWS, Microsoft et al or would not want to include references to other providers, but it would have been a useful tool for anyone going down the cloud route.
But even a service like Plan for Cloud's tool doesn't take out the human element - nothing's foolproof. Well, nothing until we learn how to stop employing humans.
