ENGINEERING

Central to the attraction of Cloud supercomputing is the “pay for only what you use” claim made by almost all service providers. But look a little deeper within the contracts and you will often find that they are not quite as attractive as first appears.

Hidden away in the small print of many cloud service contracts are a range of charges that customers may not ‘spot’ until it is too late; that is until very large bills for data processing or data storage have made their eyes bulge. The problem is compounded in that individual users, potentially junior grades, have the potential to rack up significant bills without any clear financial monitoring or control. So suppliers really need to be more transparent and customers need to be more demanding for price clarity, otherwise it will severely limit the growth of the cloud services business.

Here are some pricing examples: 

• Cloud pricing will often incur general usage costs, i.e. pay per use, but a supplier may insist on a minimum charge of 2-hours, for example, even though the customer only requires 20-minutes of compute power
• Similarly, part hours are also often rounded-up for charging back. For example, 2 hours and 5 minutes of usage would be charged at 3-hours
• Users may also find that attractive initial rates apply only up to a certain level of use, and beyond that premium rates kick in, escalating the overall cost
• Often missed by users in the small print are charges for bandwidth in / out, i.e. the transfer of data. Even if spotted, this is a cost that may be difficult to estimate and even harder to control. So be very aware of these charges 
• Suppliers can charge for the number of users accessing the resource - with costs increasing in stages, e.g. 5 users or 10 users – similar to the traditional software licensing model that most are familiar with.

None of the charges are particularly unfair. However, when these different pricing structures are combined and not transparent, it creates a very confusing picture for the customer who has the challenge of implementing controls over the use of the service if he is to deliver the benefits defined in his business case for cloud.

On a related point, data duplication can also affect cloud storage costs and should be taken seriously. It is not uncommon for multiple versions of a file to exist that are mostly identical and are not necessarily required, but are costing the user for their storage. A 2012 study from Johannes Gutenberg University Mainz, the Barcelona Supercomputing Center and University of Hamburg, on High Performance Computing [HPC] data sets, found typically 20 to 30 per cent of this online data could be removed by applying data de-duplication techniques, peaking up to 70 per cent for some data sets.

Tarnish

There is a growing concern amongst IT and finance managers over this issue. We see it with our own Compute-On-Demand service, enCORE, where prospective clients are asking searching questions about fee structures. We actually expect users to be cautious and certainly don’t want to discourage that level of wariness. Customers are using enCORE partly because of our open, honest pricing policy, but many are hearing the message from employers to take a cautious approach, to be wary of anything that sounds like it might carry the ‘cloud services’ badge.

Action

For smaller business particularly, the benefits of cloud far outweigh the pricing ‘hurdles’ of using such a service, but caveat emptor certainly applies here.

As with everything, the devil is in the detail. Right now, users should look for a supplier that will enable self-monitoring of their spend in real-time whilst they’re using a cloud service, not when the bill is drops onto the doormat. Companies must set in-house policies to control usage by users – it’s very easy to get carried away on a cloud service. Users must check the fine print of the contract and ask questions if it is to confusing to understand; look for a supplier that is prepared to explain the billing process, not rush you into signing a contract. Perhaps most importantly, pick a supplier, partner up and let them guide you through the process; plug and play isn’t always cheapest in the long run.