Our perception of the cloud is underpinned by three terms: scalable, flexible, and efficient. These adjectives, however, are only the capabilities of the cloud—not the benefits. Turning the cloud’s capabilities into benefits involves understanding the nature of the cloud and optimizing how you use it based on that.

From choosing your cloud vendor to rewriting your code for the cloud, our cloud cost efficiency checklist spans a wide spectrum of planning and actions. The questions you should be asking yourself include:

  1. Which is the right cloud provider for us?

  2. Are our resources right-sized?

  3. Are unused resources cleaned up regularly?

  4. How effectively are we leveraging savings plans?

  5. Are we making use of spot instances?

  6. How efficient is our auto-scaling configuration?

  7. Are our applications cloud-efficient?

  8. How optimized are our network pathways?

For length concerns, we have split the article into two parts, each answering four questions, with the second part dropping in a couple of weeks. You can, however, get immediate access to the full checklist by downloading our latest white paper, How IT leaders can drive more with less: An enterprise guide to technology adoption and cloud usage in a disrupted economy.

1. Which is the right cloud provider for us?  

Cloud cost optimization starts with choosing the right cloud provider and the right cloud model, whether it’s single or multi-cloud. Here are the aspects to consider first:

Your applications’ service requirements: A clear understanding of your organization’s and your industry’s service requirements is key to choosing the right cloud vendor. Each vendor comes with its own service portfolio and specializations. Choose the one that best addresses your organization’s cloud requirements.

E.g., organizations with diverse service requirements can consider AWS, known for its broad service portfolio, whereas organizations with specific business objectives with ML requirements can consider GCP, thanks to BigQuery’s data analytics prowess.

Existing technology investments and licensing agreements: There can be cost advantages when your cloud provider also offers the technology stack you already use.

E.g., if you have significantly invested in Microsoft technologies, such as Windows Server and Microsoft SQL, then managing them in Azure is more cost-efficient than in other cloud solutions.

High-volume data transfer needs: Large volumes of data may incur higher costs in a multi-cloud scenario due to data transfer charges between providers.

E.g., if your organization needs to move a substantial volume of data from storage hosted with cloud provider A to a database service with cloud provider B, the data transfer costs will be higher than they would be when moving data within the same vendor ecosystem.

2. Are our resources right-sized?

Businesses adopt cloud services, thinking they pay only for what they use. But a more helpful way of putting it would be they pay for what they provision. That speaks to the importance of right-sizing, which means selecting the right size and type of cloud resources for each workload and avoiding over- and underprovisioning resources.

Selecting instances: Compute, memory, and storage are broadly considered before selecting an instance. There are general instances where all are in balance, suitable for workloads like web servers, and then there are instances optimized for one particular metric.

E.g., a compute-optimized instance comes with the lowest price per unit of compute, which is ideal for compute-intensive applications, like games.

Choosing the right instance for your workload not only determines the performance but also the cost efficiency.

Overprovisioning: Overprovisioning is usually due to:

a. Poor workload predictions, e.g., allocating resources during new product or feature rollouts based on a certain amount of traffic. Those resources are left unutilized when the expected traffic doesn’t turn up.

b. Unoptimized resource allocation, e.g., allocating an Elastic Compute Cloud instance with a boot volume of 50GB is cost-inefficient when the OS of the applications requires a smaller percentage of disk space.

Accurately predicting workloads and workload-optimized resource allocation are key to avoid over- and underprovisioning.

3. Are unused resources cleaned up regularly?

This refers to the discipline of removing unused resources and deactivating unnecessary components in the cloud environment on a regular basis.

Unused instances: These are instances that are used for a specific period but then sit idle for a majority of the time.

E.g., VMs provisioned for once-a-year peak in workloads remain unused for the rest of the year, or VMs used in the development environment sit idle during non-business hours.

Identifying and terminating such instances will save you from accruing additional costs for unused instances. Using automation scripts is an industry best practice to tackle this.

Orphaned storage instances: Storage resources can accumulate, especially if they’re not associated with any active instances or services.

E.g., you may use automation scripts to create and delete VMs whenever a project need occurs, but you may not be aware of the storage instances being created along the way. This accumulates storage instances that don’t serve the workload but reflect in bills.

To combat this, scan for storage volumes and snapshots that are no longer in use and remove these orphaned resources.

Obsolete network components: Load balancers and networking components may become obsolete due to changes in application architecture or scaling requirements.

E.g., while modernizing applications, the load balancer configured for monolithic architecture will become obsolete as microservices are spread across servers and services.

Automation scripts can regularly audit the network topology and identify components that are no longer serving a purpose.

4. How effectively are we leveraging savings plans?  

When you have workloads running at a high scale or for long period, you can leverage cloud providers’ savings plans. Generally, they come in two types: reserving instances and volume-based discounts.

Reserving instances: The longer the contract of a cloud service, the lower the unit price. This works for application workloads you will be running for years. You can avoid paying on-demand usage price for them by identifying their baseline consumption and opting for reserved instances.

That said, since you prepay for such instances, there’s also a risk of investment loss due to underutilization. So, you need clarity into the resource allocation of a workload, which takes at least one cycle of observation.

There’s a smart way to manage this. You can opt for reserved instances for the developing and testing environment alone. Once you get familiar with the application’s usage pattern, apply the insights to the production environment.

Volume-based discounts: The larger the volume of resources, the smaller the unit price. This works for high-volume workloads.

E.g., trading institutions performing real-time data analysis on a daily basis can aim for discounted costs for the data they process and compute resources they use.

We recommend hosting such workloads from a single vendor to leverage this discount because resource volumes split across vendors are less likely to reach the discount-eligible thresholds a particular vendor offers. Single-cloud hosting, however, affects disaster recovery capabilities. So, ensure proper redundancy measures if you want to leverage volume-based discounts for critical workloads.

We will be answering the remaining four questions in the second part of this blog series, releasing in two weeks.

Meanwhile, if you’re aiming to optimize your IT budget for your digital transformation goals, explore our latest white paper, How IT leaders can drive more with less, which can help you in four ways:

1. Make effective use of the budget for standard IT requirements.

Learn how our technology adoption framework can help you allocate your IT investments strategically to obtain amplified business outcomes.

2. Conventional IT investments aren’t enough. You need a growth engine, too.

Read our case study on General Electric to learn how you can apply technology on top of existing business resources and build an alternate revenue engine.

3. When it comes to cost, the cloud is a double-edged sword. Make sure you use the right edge of it.

Learn the role of CIOs in the cloud cost optimization cycle and leverage our cloud cost efficiency checklist to master the fundamentals of cloud usage.

4. Enterprise organizations are saving 60% of their cloud costs by shifting their workloads to their own data centers.

Read the cloud exit story of a SaaS enterprise and the behind-the-scenes math of its savings in detail.