Cloud cost optimisation – best practices

When you considered migrating your business, or some of its segments on the cloud, one of the main pros was the flexibility and the pay only what you use model. You left the old-school unneeded capex behind and started enjoying the benefits of progressive cloud technology. This means complete transparency, spending control and overall cost reduction.

But only if you know exactly what you’re using. All the time.

So, you have migrated your infrastructure to the public cloud, yet, somehow, the spending didn’t go on the downline. Maybe, on the contrary, it started increasing? Why is this?

Because you’re paying for what you provision. Every cloud is a “living thing”, so it requires around the clock improvements – since there is a massive space for it. Enters cloud cost optimisation.

 “A cost-optimized system will fully utilize all resources, achieve an outcome at the lowest possible price point, and meet your functional requirements”

The challenge of managing cloud spend grows as cloud use increases. According to studies, cost optimisation was the top priority for cloud users in 2019.

Setting a cost optimisation strategy

Cloud cost optimisation is an ongoing process that requires time, knowledge and devotion, but comes with amazing results. If you want to see your infrastructure on cloud perfectly designed and set to decrease the cost while increasing the productivity, you need to start with setting a strategy that will include cost ownership, deployment processes and reporting.

To begin with, your business model and internal organization must be adjusted to your new technology. Meaning that, across your organization, the teams should be educated on how the cloud works and which steps to take to ensure maximum utilization and minimum or no waste. E.g. the finance team will, of course, want to keep track of the spending while C-level needs to understand the impact of the cloud cost and its’ ROI in the long-run.

When you have set the grounds of optimal cost management, you can start putting the cloud spending puzzle together.

Identify unused and idle instances

Believe it or not, this scenario is quite often – the administrator or the developer used some server for a function testing and forgot to detach it. The instance continues to pile up costs with no actual usage. On the other hand, some of the instances are actually being used, but only around 15% or so of the capacity. The bill will, however, show 100% utilization. Take these situations seriously, as the unwanted cost can grow dramatically and provoke overspending.

Identifying the unused and complete reorganization of active instances requires time, knowledge and regular reporting. Consulting cloud management experts is always a good idea if you don’t have inhouse resources to proceed with this venture.

Right-sizing

Right-sizing implies ultimate consumption balance – the problem is, it’s easier said than done. With literally millions of instances to choose from, a cloud administrator has a complex job ahead of him. One which is meaningless if there is no analysis of resource usage or chargebacks enabled. Providers like AWS and Azure offer various tools to help you diagnose the optimal infrastructure and decide on the next steps. Whether you decide to check it out yourself or seek help from cloud experts, bear in mind that rightsizing does not only decrease cloud costs, it helps the whole architecture run smoothly and efficiently.

Try the tools: Heatmaps

Even if you are a 365/24/7 business, you still have peaks and lows and your cloud environment should feel that pulse and respond accordingly. This is why heatmaps are such a beloved tool. Heatmaps enable you to track the levels of cloud computing demand and create adequate usage patterns. For example, some of your instances are used only for testing and staging – so they can be turned off after work and on weekends and run full power Monday to Friday during working hours.

Think twice on the multi-cloud

Depending on your needs and capacities, multi-cloud can be just the right or perfectly wrong strategy for your business. Some companies choose multi-cloud as they refuse to be locked into one vendor. Others want to ensure the uptime and availability are impeccable.

But when it comes to cost management, you might want to put to paper your monthly costs per vendor, add that and check whether this spending may mean a significant volume discount given to a single cloud provider. If you can, make sure to consult cloud experts who can clearly see your position and future ambitions and help you craft the best (multi) cloud cost strategy.

Save with reserved instances

Most cloud providers offer so-called reserved instances. Reserved instances are not some kind of special machine types – they are, basically, a contract which says that, in exchange of a significant discount, you’re committing yourself to pay for the reserved instance for a specified time, usually one or more years. Pay attention that we said ‘pay’ not ‘use’ – you’ll pay for your reserved instance for one year, even if you use it for only 10 months. However, since the reserved instance is just preselling a certain amount of resources, some cloud providers offer the flexibility of how the consumption of those resources is calculated.

For instance, you can buy a 4 vCPU, 16 GB RAM virtual machine reserved instance for one year, but that doesn’t mean that you actually need to use such a virtual machine. You can instantiate two 2 vCPU 8GB RAM virtual machines for 10 months and then for last two months you delete one of those virtual machines and instantiate one with 4 vCPU 16 GB RAM virtual machine. In that case, you’ll pay for your reserved instance for one year and an equivalent of one on-demand 2vCPU 8GB RAM virtual machine instance for 2 months. However, be careful of the gotchas in how your cloud provider calculates the usage of reserved instances.

If you are using Microsoft Azure, Azure Reserved Instances (Azure RI) are among options to save your cloud cost. By committing a one-year or three-year contract for a virtual machine in Azure, you can save up to 72% compared to the pay-as-you-go model. Of course, there are a lot of pitfalls in terms of reservation, so you need to check what can be reserved and what are the conditions. For example, some virtual machine size series, such as A or G, cannot be reserved, as well as any virtual machines in specific regions. Also, you need to consider whether reservations are appropriate for all of your environments, sizing, operating systems. Another important thing in terms of the reservation is a cancellation of the reservation. Azure has a very strict policy regarding this, so if you want to cancel your reservation, you will be charged for some penalty.

Play it safe with spot instances

Some cloud providers offer a possibility of limiting your costs by using the so-called spot instances. Every provider has some amount of unused resources and they aim to have near 100% utilisation. Since the amount of utilized resources changes literally each second, it is possible to set up a floating price for them which follows the basic law of supply and demand. If the demand for a specific resource rises the number of unallocated units of this resource decreases and its price rises. If the demand decreases, there is a larger percentage of free resources of that kind and its price decreases. To increase the utilisation of such resource, the cloud provider offers you the possibility to allocate and use those free resources only when its price drops below some threshold. This is the spot instance. It can be instantiated automatically as soon as the price drops below the threshold, and it will run as long as the price stays below the threshold. As soon as the price rises above the threshold, your spot instances will be automatically shut down until the price goes down again.

Spot instances are not for every usage scenario. You cannot predict when they will be available nor for how long your instances will be running. This means that spot instances are good for some batch jobs that are not time-restricted and which execution could be stopped and restarted at any moment.

As you can see, adopting the cloud means that you are about to embark on a dynamic and complex journey to achieve business prosperity on many levels. There is a reason why the vast majority of companies are spending most of their efforts to reduce costs on the cloud and improve the overall efficiency. And for many of them, the advice is simple: begin your cloud cost optimisation with a cost analysis and clear goal setting, and then train your staff to embrace the required needs and steps.

If you wish to hear more about the options and opportunities from an expert, consultations with a SuperAdmins Cloud Solutions Architect are free of charge.