There are two aspects to cloud savings. First, using cloud computing can in itself generate cost savings for organizations. Indeed, cost savings are a primary reason many organizations are moving resources to the cloud.
Second, once already in the cloud, organizations need to control and optimize cloud costs. Strategies for reducing cloud costs include identifying poorly managed resources, eliminating waste, achieving larger discounts by leveraging pricing models like reserved instances and spot instances, and scaling cloud resources appropriately to match actual demand.
A key point to understand about cloud costs is that, while the promise of the cloud is to bill for actual resources used, in reality most cloud services bill you for resources even if you don’t fully utilize them. A Gartner report estimates that as much as 70% of cloud costs are wasted. Most strategies for reducing cloud costs involve gaining visibility into cloud resources, identifying waste and eliminating it.
In this article, you will learn:
Cost savings was, and continues to be, a main driver of cloud adoption. Cloud computing can save organizations costs in a variety of ways. Let's look at the most common of these.
When migrating infrastructure to the cloud, the cloud provider takes responsibility for the operation of services and equipment under a predefined service level agreement (SLA). For those areas that the cloud provider takes over, organizations can save on overhead costs, as they no longer need to invest time monitoring, securing, and maintaining computing systems.
Keep in mind that all cloud providers use a shared responsibility model, where the cloud provider operates one part of the technology stack (for example, the underlying servers) and the customer is responsible for another part (for example, workloads and data). Your organization will only save on operating costs for those elements that are managed by the cloud provider.
In the past, organizations had to plan for peak performance and purchase hardware equipment that could meet the largest possible load expected by their applications. This could lead to underutilized or completely unused hardware. With cloud computing, almost any compute load can be provisioned on demand, scaled up according to actual demand, and scaled down or shut down completely when not needed.
This means that IT capital costs can now be converted into operating costs. In the short term, this substantially reduces capital expenditure. In the longer term, each organization must evaluate the long term cost of computing, when consuming it as a cloud service vs. buying equipment on-premises—in many cases, the cloud may actually be more expensive.
An organization’s ability to make efficient use of resources, shifting them to where they are needed most, is also a cost saver.
In an on-premises data center, organizations have limited ability to adjust to changing business needs and ensure maximal utilization of resources. In the cloud, the wide variety of services,types of infrastructure available, advanced automation capabilities, and wide range of available geographic locations, makes it possible to immediately react to almost any business need. This reduces waste and improves the return on investment of IT expenditure.
For any business, especially for large organizations, downtime causes major damage in terms of lost revenue, productivity, and reputation. The cloud makes computing services much more resilient than they could ever be in an on-premises data center.
All cloud providers let you run resources on multiple data centers in separate physical locations. It is typically much easier to backup and restore systems in the cloud, and it is possible to set up remote disaster recovery sites at the click of a button (whereas in a local data center this would require a massive investment). By improving business continuity, businesses can save on the expected cost of downtime.
Following are several strategies, which, when followed consistently, can generate cost savings in the cloud for almost any organization.
Cloud providers offer discount plans in exchange for a long-term commitment to cloud resources, including enterprise agreements, savings plans and reserved instances.
However, be careful before you commit, because these discounts will lock you into specific cloud services, instances, or resource configurations, which may be suitable for your current needs, but may not be appropriate further down the line. Because reserved instances lock you into a specific resource for a term of 1 or 3 years, this can create major inefficiencies.
In most cases, you can generate significant savings by optimizing existing resources (e.g. right-sizing underutilized resources). Start with an assessment that can determine how much you can save without any long-term commitment. Then, only at a second stage, consider long-term discount options.
If you commit to discounted pricing models first, many optimization options may no longer be available (for example, you may not be able to remove or down-size underutilized instances, because you have already committed to them for a year or more).
Cloud providers offer several discounted pricing models that can generate significant price savings. However, as mentioned above, it is advised to first investigate short-term optimization options, before considering long-term commitments like reserved instances and savings plans.
All cloud providers offer capacity reservations, typically known as reserved instances (RI). While different providers offer different terms for RIs, they can typically save you around 50-70% compared to on-demand rates for the same cloud resources.
Generally, when applying this option, a discount is provided based on the term of the contract (one or three years) and the amount of the upfront payment. Cloud providers and third-party tools offer automated ways to recommend which instances should be converted to RIs (based on the nature of workloads running on them), and the potential saving opportunities.
It is recommended to use RIs for continuously running cloud workloads with stable usage patterns. Use the minimum amount of RIs possible—any dynamic capacity,, should be handled using on-demand instances or other pricing options. It is also a good idea to consolidate RI purchases into one cloud account, to give your team better visibility of these resources, and ensure they are used at all times.
Learn more in our detailed guide to purchasing RIs on cloud marketplaces
AWS Savings Plans is a flexible pricing model that offers a lower price than the on-demand price in exchange for a one- or three-year contract of use (calculated in terms of spend/hour). AWS offers three categories of Savings Plans: Compute Savings Plans, EC2 Savings Plans and Machine Learning Savings Plans (SageMaker Savings Plans).
Savings Plans can save you a lot of money in exchange for a contract of use for one or three years in a row. For example, the EC2 Instance Savings Plan can save up to 72% of your Amazon EC2 instance usage when compared to on-demand pricing. The Amazon SageMaker Savings Plan can save up to 64% of your Amazon SageMaker service usage.
Spot instances are spare computing capacity which cloud providers must maintain in the event of surges in demand. Pricing can be significantly lower than on-demand rates, with discounts up to 90%. However, it can be tricky to make use of spot instances, because the cloud provider can terminate them with short notice, as soon as capacity is required for on-demand or reserved capacity customers.
Because spot instances can be terminated at any time, they are commonly used for workloads that can tolerate failure—like stateless microservices, batch operations, and fault tolerant services.
For mission-critical tasks or tasks that cannot be stopped and restarted, spot instances are more difficult to use. However, with the aid of automated tools such as Elastigroup from Spot by NetApp, you can manage spot instances to maximize availability, making them relevant for mission critical workloads as well.
Another saving option is burstable instances—these are low-cost compute instances designed to provide the lowest level of CPU performance, but with the ability to “burst” to a higher level of performance if requested by the workload. Burst instances use the concept of CPU credits—when running at the low level of performance, they accumulate credits, and when they need to “burst”, they spend credits.
Burstable instances are ideal for a variety of general purpose applications with low CPU usage and occasional peaks in CPU utilization. Examples include small and medium-sized databases, smaller-scale websites and web applications, and dev/test environments.
Learn more in our detailed guide to cloud cost models
The goal of cloud cost optimization is to minimize costs by eliminating overused or underutilized computing resources, and ensuring exactly the right number of resources are available current loads. Collectively, this is known as “rightsizing”.
To rightsize cloud resources, you should first analyze utilization on existing virtual machines, decide what changes need to be made to improve efficiency, allocate the appropriate resources, and tune your infrastructure to make sure you are getting the best performance out of your existing assets.
Cloud providers provide instrumentation and monitoring tools, allowing you to track memory, CPU, and network usage on all types of cloud resources. This allows you to determine resource consumption and define upper and lower thresholds. It is best to automate your response to high or low utilization—for example, scaling up when utilization is too high, or scaling down when it is too low.
Tools like Cloud Analyzer from Spot by NetApp can help you assess utilization across your cloud environment, and generate smart right sizing recommendations, to ensure you are only paying for the resources you actually need.
Cloud governance is often a complex undertaking. To ensure efficiency and reduce costs, organizations are adding automation to their tooling stack. Particularly, automated cloud policies are highly useful in keeping track of resources and reducing costs. Here are several examples of policies that can help reduce costs in your cloud infrastructure:
Related content: read our guide to cloud cost optimization (coming soon)
Software licenses are often limited to use in certain environments. This means that if you are currently using licenses that were purchased for your on-premise infrastructure, you may not be able to use the same licenses in the cloud. You need to create an inventory of licenses, understand the restrictions, and then determine which licenses are most appropriate for your budget.
In some cases, you might be able to reduce licensing costs by leveraging a cloud service, like Azure Hybrid Benefit, which helps you save costs by letting you bring your own license (BYOL) to the cloud. However, Oracle databases come with license restrictions that might make BYOL too expensive to implement.
Going beyond cloud analytics and recommendations, Spot by NetApp automates and optimizes your cloud infrastructure in AWS, Azure or Google Cloud to deliver SLA-backed availability and performance at the lowest possible cost.
Our technology uses machine learning and analytics to effortlessly and affordably scale any workload, from stateful single instances to cloud-native clusters, using the ideal mix of cloud compute resources, from on-demand to reserved and spot instances.
for up to 20 instances