In today's digital world, companies heavily depend on cloud solutions for their business operations. However, as the reliance on cloud services grows, so does the cost. Some organisations are so concerned with their increasing bills that they are bucking the trend and retrenching back to on-premise data centres. In this article, we explore a strategic approach to optimising cloud costs that can lead to a fundamental change in the way organisations manage their cloud spend.
Cost optimisation is a way of achieving efficiency that can take the form of reduced costs or increased value. Businesses survive by doing the same for less but thrive by doing more for the same. Organisations need to systematically ensure that every pound, dollar or euro spent in the cloud delivers the maximum value to the organisation and its customers. Few organisations can afford to waste money and, surprisingly, a more systematic approach to cost optimisation is that common.
CHALLENGES
Cloud costs have always been contentious. A large part of on-premise operations are covered by hidden sunk costs in data centre buildings, cooling systems and racking; and already accepted overheads such as electricity and data centre staff. This leaves the simple, predictable, transparent but artificially low cost of server hardware payable by each project. Some organisations carefully apportion sunk costs and overhead costs between projects but that seems to be the exception rather than the rule.
In contrast, cloud computing costs are transparent and directly tied to usage: every request, stored byte of data or moving byte of data contributes to a bill under one or more pricing plans. Whilst cloud costs are no longer hidden, they come with their own set of challenges. Tools do exist to help forecast spend, but they require expert knowledge to understand the costs incurred and the rate of spend. Cloud costs tend to be visited at the start of projects to provide a high-level estimate of operating costs but cost is rarely considered after that. That is until the unexpectedly high bill arrives!
Cloud can often be seen as more expensive than an on-premise solution. However, a fair comparison should take into account the sunk costs and overheads of an on-premises solution, as well as the cost of up-front purchase of hardware. Any true comparison should consider on-premise sunk costs and overheads, the cost of cash flow incurred by the upfront purchase of hardware as well as the more obvious costs. In such a comparison, the cloud is found to be more capable and flexible, delivering more value for a similar cost. Examples of capability and flexibility include elastic scaling, near-infinite resources and on-demand access to the latest services, including artificial intelligence. Simply put, cloud computing offers better value for money.
However, getting cloud costs under control has been a significant problem for many organisations often stemming from a chaotic adoption of cloud. In many organisations, infrastructure teams were tasked with managing the adoption of the cloud. In the absence of established industry best practices, these teams followed on-premise style processes. They delivered cloud solutions with the same steady cadence of on-premise solutions. To speed up deployment and leverage more of the cloud’s capabilities, development teams took up the challenge. They were faster and more innovative in their use of the cloud’s capabilities but with limited or no regard for cost management that infrastructure teams had often been better with. The result has been the cost bomb that has been exploding on organisations’ bottom lines.
As a result, organisations today are faced with high cloud costs that are complex, difficult to understand and difficult to predict. What is worse, the very capability that has the potential to deliver significant cost savings and better value for money is seen as expensive and in some cases being shunned.
FUNDAMENTALS
The road to effective cloud cost management starts with understanding the fundamentals:
Use it wisely: don’t turn it on until you need it - most services in the cloud are billed by the minute and costs are not incurred until they are used. This contrasts favourably with on-premise solutions which are paid for the lifetime of the server upfront.
Use it efficiently - when services are being used, ensure the most value is being obtained for what is being billed. Part of what makes cloud billing complex is that different services are billed for different metrics such as the number of requests, number of bytes etc. Many are charged against multiple metrics, and customers are charged per request and for the number of bytes. Rightsizing the solution reduces waste and improves efficiency. Again, this contrasts with on-premise solutions which have fixed size and are typically over-specified initially and later become insufficient to meet demand.
Turn it off when finished - unsurprisingly, costs can be saved by turning a service off once it is finished … unless it will be reused shortly after. Even the most complex cloud capabilities can be provisioned in a few minutes. If a service is not required for the weekend, over a holiday or when not testing, it can be turned off.
That’s it. It really isn’t any more complex at the base level than that. How this is achieved, is a little more involved.
MONITOR, PRIORITISE, TECHNOLOGY AND ITERATE
The first step in optimising cloud costs is to implement proper monitoring and reporting. Each product or system must report its costs on at least a monthly basis.
Many organisations fail to budget cloud costs. The report can be used to baseline monthly spending and then future months may be assessed against that baseline. Any change in costs should be understood and documented, providing a clear indication of which decisions are driving which cost changes. It supports engineers, product and project managers in understanding the cost implications of their actions and encourages a culture of joint responsibility for cost. To be clear, the product or system owner remains accountable for the cost, but every member of the team is responsible for understanding the costs and their impact.
After understanding the costs, the next step is to prioritise. For most implementations, the largest costs are present in the recurring monthly bill. However, for some implementations, one-off costs such as imports or migrations can generate the highest expenses. In either case, the focus should be on the most significant cost items.
When prioritising, the first question to ask is: 'Is it necessary?'. Many organisations continue to run zombie systems with no users due to a lack of product or system ownership. If it is necessary, the solution should be assessed from the top down. The architecture has a much greater impact on the cost than any specific line of code (although there are exceptions). A well-implemented bad architecture implemented well will generally cost more to operate than a badly implemented good architecture. The latter will also be easier to fix.
The architecture should be checked to ensure that it is making appropriate use of technology, avoiding unnecessary duplication and taking advantage of cloud capabilities that can control costs. Consider the following areas, although a complete checklist is beyond the scope of this article:
Compute: dynamic scaling can significantly reduce costs by allowing solutions to react to changes in demand. Serverless computing is often good for dynamic scaling with the amazing ability to scale down to nothing in the absence of demand. Compute resources should be scaled separately matching memory and processor specifications for the task. It may also make sense to separate tasks that have widely different hardware needs avoiding overspecified resources.
Storage: not all storage is the same. Storage temperature (hot vs cold) and redundancy have the biggest impact on cost. Not every part of a solution needs three or more copies of data on the fastest disks available. Clouds also offer comprehensive logging capabilities. In a restaurant, you don’t normally order every item on the menu. Similarly, you don’t need to log every event or keep the logs indefinitely. Selecting the right level of logging provides invaluable insight whilst minimising clutter in the logs and managing costs.
Networking: be aware of the cost of moving data. All the main cloud providers offer free ingress for data but charge for egress. These charges can be significant and are all too often overlooked in multi-cloud strategies. Services within clouds sometimes charge for bytes entering or leaving the service. A sensible architecture moves data only when it is needed.
Reserve and spot instances: on-demand computing is one of the hallmarks of cloud computing. It supports experimentation and changes in direction. However, for most enterprises, a significant proportion of usage is predictable a long way in advance. For these workloads, reserved instances provide predictability for cloud providers who in turn can pass cost savings back to customers. Spot instances also offer significant cost savings but most organisations' workloads will not easily match resources that may or may not be available and could be removed at short notice.
Expensive services: some cloud services are expensive. This may often be justified by the unique capabilities they offer. However, unless these capabilities are not being used, it is often better to consider alternative options. One example is API Management. For many applications, this provides a sensible abstraction of needed capabilities including authorisation, authentication, caching, rate limiting and routing. However, some internal APIs do not need all these capabilities and it is often worth examining the value for money of employing these services.
None of these changes should be done without forecasting the cost impact. It is important to test understanding and build the teams’ experience and confidence in predicting costs. A good way to start is with one of the many cloud pricing calculators which all the main vendors provide. This allows you to describe the services you are considering and work out the cost. For existing solutions, this should match the existing monthly costs. If it doesn’t match, something is different. Once they match changes in compute or storage can be forecast before the change is made. It also spotlights all costs, including those often overlooked such as egress charges.
Having monitored, understood the costs, focused on the larger, priority areas, examined the architecture, predicted and made the changes, it is time to loop back to the start. Monitor to confirm that the changes are delivering the expected cost savings. Some small organisations have halved their cloud costs on the first few iterations of cost optimisation. Future iterations will yield less savings but they are by no means less important.
Embedding a culture of continuous cost optimisation in an organisation will drive better decision-making at the start of projects with far more cost-effective solutions going into production. Central to this is a culture of shared responsibility for cloud costs. If your organisation has a cloud cost problem, ask yourself if there is a culture of shared responsibility. Chances are that no one believes it is their responsibility and the person accountable has insufficient detailed knowledge to know what to do about it.
Starting to create a culture of shared responsibility for cost is as simple as:
Monitoring all current cloud usage
Reporting on all usage by project or system, at least monthly
Understanding the costs and impact of decisions on those costs
Prioritising the biggest consumers of cost
Forecast cost changes
Optimise costs by making the necessary changes
Iterating round to monitor, report, understand, prioritise, forecast and optimise continuously
START YOUR JOURNEY TO CLOUD COST EFFICIENCY
Effective cloud cost management isn't just about cutting costs; it's about maximising value and ensuring your resources are used wisely. By embedding a culture of continuous cost optimisation, your organisation can make better decisions from the outset, leading to more cost-effective solutions and significant savings over time. Start your journey to cloud cost efficiency today.
Ready to transform your cloud strategy and achieve true cost optimisation? Contact our team for a personalised consultation and take the first step towards a more efficient, cost-effective cloud infrastructure, let’s talk!
Comments