How can organisations optimise costs for public cloud platforms?

Many organisations look for cost savings as a consequence of using public cloud, so controlling cost is important to any business in the process of migrating to the cloud. As delivery teams’ decisions directly impact the costs of their services, they become responsible not only for operating their applications and services, but also for playing a direct role in managing costs. With this in mind, moving to the cloud requires not only a technology shift, but also a cultural shift to a more cost-conscious cloud culture within organisations.

In this article, we outline how companies can help manage and reduce costs when moving to the cloud.

Read next: Whitepaper:  Cloud optimisation best practices: Finding your cloud's silver lining

Cost considerations when using the public cloud

Cost optimisation enables companies to save money that can then be re-invested in more valuable innovation. The key areas that we will be examining include:

• Matching demand and pricing: Matching performance and capacity with correct pricing;

• Cloud service model: Choosing the right “as a service model” (Iaas, Paas, or Saas);

• Continuous cost improvement: Monitoring usage and costs and driving iterative improvements.

Demand matching

Different workloads and performance patterns map best to different pricing models. Here we consider workload patterns, and when it may be appropriate to move to reserved capacity, pay-as-you-go, or pre-emptible capacity pricing.

Workload patterns

The easiest way to control cost on a cloud platform is to make sure that systems are correctly provisioned for the required performance and that they aren’t incurring extra cost for resources that are largely idle. Workload patterns do change over time and so it is important that performance is continually monitored and maintained as platform demand increases. If more performance is required, then additional compute can be added with a predictable increase in cost. Baseline workloads benefit from committing to reserved capacity, pricing up front for the largest discounts.

In connection with ensuring that platforms are provisioned with the right amount of compute power is the concept of capacity-based optimisation. Some workloads are greater at certain times of the day, month, season, or year, and thus elastic compute platforms should be leveraged. These can be auto-scaled up or down to meet capacity demand, but can also introduce variable cost. It is important to articulate these costs to budget holders earlier on so that they can plan accordingly. Variable loads can also benefit from pay-as-you-go pricing or spot / preemptible capacity.

The concept of ephemeral computing means using temporary environments only as necessary. A cloud operating platform enables us to define infrastructure as code, automate provisioning, and manage deployments with DevOps tooling. Ephemeral compute is typically used in development and test environments and can be created on demand or destroyed once the work is complete, minimising costs for the business.

Some workloads do inevitably require third-party products to be deployed on cloud infrastructure. Many software vendors have license terms that permit deploying their software to cloud on advantageous terms. Companies should work with their licensing providers to understand what Bring-Your-Own-License offers might be available to them in order to yield an additional discount.

Reserved capacity

Savings can be achieved for predictable static workloads. Each of the major cloud vendors offer their customers the option to pre-pay for an amount of reserved compute time, and the greater the committed period is, the greater the discount percentage will be. Companies should consider using the reserved compute models for services that run 24 hours a day, 365 days a year.

Companies should also consider the potential savings before switching pricing models. Reserved compute models do include a refund or cancellation policy, but the saving made may be negated if this is invoked. Each cloud vendor offer can vary, and we can help you to gain a better understanding of your options.

Pay As You Go (PAYG)

When moving to public cloud, the most common pricing model that companies initially adopt is ‘pay-as-you-go’ (PAYG). This is often a great choice for companies creating their first workloads, but as the cloud platform continues to grow, there are more cost-effective pricing models that should be considered. Environments such as development and test platforms that can be shut down when not in use may benefit from remaining on the PAYG model.

Spot or pre-emptible capacity

The major cloud vendors sell their spare capacity at a large discount as ‘spot’ or ‘pre-emptible’ instances – resources that can be used cheaply but may be reclaimed by the vendor at short notice. For variable workloads that are resource-hungry but not particularly time-critical, spot or pre-emptible instances can generate substantial savings.

Operating models - IaaS vs PaaS and SaaS in relation to cost

Cloud providers are adding more PaaS (Platform as a Service) and SaaS (Software as a Service) products to their offerings, potentially offering better value for money over a migration to basic IaaS (Infrastructure as a Service). This can help organisations to retain a lean IT footprint.

When companies first move to the cloud, the most commonly adopted operating model is Infrastructure-as-a-Service (IaaS). This often offers companies the least path of resistance to the cloud from an operational standpoint. As a cloud platform matures, it is typical for a company to re-architect their applications and services to use PaaS-based services such as container platforms. The amount of operational support the platform requires is lower and therefore operational staff can pivot to work on intellectual property or revenue generating workstreams. As companies transition their applications and services to a SaaS model, the operational responsibility lowers and this can create additional savings.

As figure 1 demonstrates, platform and administration costs are higher when operating in the on-premises or IaaS models, but lowered when PaaS and SaaS models are adopted. However, it is important to consider some of the risks to adopting PaaS and SaaS-based offerings, including the increased risk of being locked to a particular vendor platform. At the same time, multi-cloud disaster recovery solutions are often not possible when using a PaaS-based model within a multi-cloud environment.

Continuous cost management

Each of the big three cloud providers offer excellent cost visibility tooling that includes historical reporting, enabling a company to track cost increases or decreases over time. These reporting solutions are largely out of the box, but include the flexibility to categorise and label your workloads and undertake detailed analysis by importing data into analytics tools. Companies should focus on measuring spend against a KPI such as total revenue, subscribers, or orders completed to add context to the spend and change the mindset from one of cloud ‘cost' to ‘return on investment' (ROI). 

The reporting capabilities of cloud platforms allow companies to gain far greater insight into the performance of their workloads and make adjustments to their compute layer accordingly to ensure that they aren’t incurring extra costs.

Furthermore, Cloud vendors support the setup of cost-based alerting. IT administrators and cost centre owners can be made aware when cost quotas are reached, and automated actions can be configured to shut down, pause, or even destroy IT platforms as appropriate.

In a nutshell

Having control over cloud platform costs is essential. Companies should continually monitor performance and capacity to ensure that applications and services operate as expected, without overspending. Consider the use of ephemeral compute where it is appropriate, and review your operating model and re-architect applications where possible to leverage a cost benefit, whilst balancing the risk of vendor lock-in. We recommend carefully reviewing your platform usage and working with licensing providers and professional services companies to ensure that you are taking advantage of the latest pricing offers.

At Credera, we have worked with a number of our clients to optimise the cost of their compute and IT platforms both from a technology and Target Operating Model (TOM) perspective.We achieve this by evaluating the following aspects of IT platforms and company operating models:

• Architecture

• Vendor specific offers and pricing

• Ecosystem performance

• Culture