Are you spending too much on public cloud?

Foresight is a wonderful thing

Public cloud services may be a commodity purchase these days but picking compute and store resources still calls for a lot of skill and foresight.

For one thing, moving workloads from on-premises or even private cloud infrastructure is a complex process. For another, public cloud technologies are constantly evolving – with new options and services offered almost weekly. Unless you are monitoring them full-time (which is, after all, not your core business as it ours!) you can miss some tricks.

Here are some of the ways many organisations are inadvertently paying for unnecessary cloud resources – and squandering precious operational budgets as a result!

Overprovisioning and underutilisation

Sometimes it seems safer to select the largest size by default, sometimes you do it by mistake, and sometimes your needs have changed. Whatever the reason, you may well be paying for more resources than you actually need.

There is no ‘one size fits all’ about cloud. Scrutiny of your usage data should indicate by how much you are underutilising storage volumes, data warehouses and relational databases, for example. Underutilised instances are another common ‘gotcha’, as are idle hosted caching tools and inefficient containerisation which could be consolidated or even turned off.

Similarly, you may have purchased more costly resources than you need. Some hosting regions are more expensive than others, and moving resources is an option for reducing ongoing fees. Additionally, not all public cloud providers are equal – so avoid getting locked in (even by lethargy or sheer lack of headspace) when another might become more cost effective down the track.

Less-than-optimal pricing and reserved capacity

Cloud providers now offer a range of options that can offer significant savings. However, if you selected ‘on demand’ by default, you could well be spending much more than you need! Consistent workloads can be run for less on reserved instances (AWS or Azure) or committed use discounts on Google Public Cloud. Workloads requiring dynamic demand can benefit by being placed in scale groups, and batch processing loads can be run at discounted low-priority rates.

Conversely, if you started out by being cautious and opting for lower-cost reserved instances and committed use discounts, you may find that they are not actually repaying your choice.

Keep a watch for expiring reservations too; you need to review them regularly and act if workloads are about to revert to costlier standard options.

Unnecessary resources

Your cloud infrastructure represents a lot of moving parts – and it’s easy to miss if some of the resources you are paying good money for are actually no longer required.

‘Orphans’ are a common form of waste: snapshots of decommissioned systems, volumes, load balancers with no instances, unused machine images or object storage among many. This happens when a virtual machine is terminated, but the resources attached to it continue running – at a cost. Similarly, still attached but idle resources are another common cause of unnecessary spend.

Another example of why it’s so important to be aware of all your options – not just the ones you initially purchased – is getting trapped on legacy resource types. It’s a bit like your electricity bill. What seemed the best deal when you researched suppliers carefully a year ago no longer is. Now others are enjoying the benefits of more recent research, while you’re stuck paying the old higher rate.

How to keep up to date?

So, yes, public cloud has become a competitive market, but it’s still easy to pay more than you actually need. We’ve developed a Cloud Waste Checklist for cloud cost optimisation – 24 quick-fire tips covering areas of potential waste and where your public cloud settings and workloads can be tweaked to result in immediate savings.