sommai - Fotolia

Cloud storage TCO: Key pitfalls and how to avoid them

Cloud’s flexibility is part of the attraction, but it can also lead to costs that spiral. We look at key elements in cloud TCO and how to intelligently make use of the cloud

Cost is often the key factor when it comes to moving IT to the cloud. In practice, working out the cost of moving a workload or business process to the cloud is complex.

And, just as with on-premise hardware, there is more to the total cost of ownership of cloud systems than the “ticket price” alone. Cloud computing’s flexibility and consumption-based pricing models can make it hard to predict the lifetime cost of a project or a system. It is easy to add capacity or performance, but it is also easy to incur additional costs.

In some areas, however, costs have become more predictable as the industry has matured. Egress costs to take data out of cloud storage is often thought of as a “hidden cost” of cloud, and certainly one that is disliked by chief information officers and service users.

But, in practice, cloud service providers are open about such costs. They can be planned for, and service consumption designed to minimise them. Firms have become more mature in their use of the cloud and are better able to use it in the most efficient way. That includes using “cold” storage for long-term archiving, but not for data that is frequently re-accessed. Or, by making use of the cloud’s unique features, rather than shifting existing workloads to the cloud wholesale.

Enterprises currently run 30% to 40% of their workloads in the cloud, says Adrian Bradley, head of cloud transformation at KPMG – but he describes their journey so far as “relatively messy”.

“When people go to the cloud, it is really tempting to manage migrations on the basis of throughputs rather than quality,” he says. “Throughput is nice and easy to measure, but when you get to the cloud, you’re much more likely to ask questions around whether you’re getting value out of it.

Instead, firms are looking for a greater degree of IT transformation through the cloud, and this is prompting them to look again at the total cost of ownership.

TCO, cloud and on-premise costs

Total cost of ownership (TCO) in the cloud brings together all operational costs associated with the task or business process.

At one level, a basic cost figure for the cloud is relatively simple to calculate. Cloud companies list their prices for storage, compute and other core costs. These are an all-in price.

Customers that buy cloud capacity do not have to factor in the cost of labour, security, IT systems management, or power, cooling and property costs. These are all budget items that have to be considered for on-premise systems and are rolled into the per-unit cost.

Importantly, firms also avoid the need for capital spending on hardware, and are instead able to charge cloud consumption fees to the operational, or Opex, budget. This removes financing costs and frees up capital to be invested elsewhere.

These are all positives, but not all costs can be eliminated entirely. IT departments still need staff to operate and manage the cloud environment. As KPMG’s Bradley points out, cloud operations are often run by more highly skilled, more costly staff than on-premise operations, and cloud specialists are less likely to be offshore.

Customers need tools to manage their cloud estate. Security and backup and recovery will also be needed. Most of those costs will be shared with on-premise systems, and may fall as the percentage of cloud systems increases.

But it is wrong to think the cloud will eliminate all on-premise costs, especially labour. In fact, organisations might face higher costs in some areas, especially if they need to hire people with specialist cloud expertise that commands a premium. “It’s really hard to get granular breakdowns on who spends how much time doing a function, and who that should be charged to,” cautions Tony Lock of analyst Freeform Dynamics.

In addition, firms need to allow for migration costs. Applications might require different licences for cloud use, and may need modification, rewriting or even a ground-up redesign. This will add to labour costs, either directly or through a systems integrator, consulting firm, or the cloud or software company’s own professional services.

Lastly, “hidden” cloud costs have not completely gone away. Costs such as data egress, where firms pay to retrieve, repatriate or even simply copy data to another application, are much better known than they were. But they are still a factor.

So too are data duplication costs, especially if the business needs to keep multiple copies of data or run compute in different availability zones that may vary in cost for the same services.

Cloud services can also offer variable pricing, with discounts for pre-booked capacity (rather than on-demand usage) and discounts for off-peak usage.

Providers also typically offer their best pricing to customers who agree to take a fixed volume of services up-front. This can be significantly cheaper than “spot” charges, but it makes TCO harder to calculate, and can encourage the over-provisioning of cloud services.

“Ingress and egress costs are important, but generally, I think they’re quite well understood,” says KPMG’s Bradley. “But what we often see is that enterprises who use cloud consume a lot more than they expect.”

Planning for cloud costs

Cost is no longer the main driver for moving to the cloud, if it ever was.

“Cost is an important factor, but it’s not usually the primary reason for doing anything,” says Freeform Dynamics’ Lock. And, he says, this has been the case for 10 to 15 years now. “That saving costs thing has always been hype.”

But if firms are looking for value from the cloud, they need to understand its lifetime costs.

Over-provisioning, or over-consumption, is probably the greatest financial risk of moving to the cloud because of its ability to scale.

Research by HashiCorp, a software supplier, found that 20% to 40% of organisations’ cloud spend is on “over-provisioned, unused and orphaned infrastructure”, simply because it is so easy to move workloads and data to the cloud.

A clearer understanding of the total costs of cloud technology, and how that compares with on-premise systems – with all their (unavoidable) overheads – is essential to avoid waste and make the most of the cloud.

Read more on Datacentre capacity planning