You are here

How to optimize your cloud spend in a multi-cloud world

public://webform/writeforus/profile-pictures/vijay.jpg
Vijay Rayapati, General Manager, Nutanix Beam

Affordability is second only to agility on the list of reasons enterprises consume cloud services. Even with pay-as-you-go pricing, however, you're probably spending too much.

Enterprises are in the honeymoon phase with cloud computing. They love the agility they get with as-a-service options—spinning up resources as needed without the lengthy and expensive process of buying, installing, and testing systems prior to production.

By allowing you to avoid big, up-front capital expenditures and letting you pay as you go, SaaS, IaaS, and PaaS have become a boon to corporate budgets. For the most part, anyway. Cloud is fast, easy, and agile, but that doesn't necessarily mean it's cheap—particularly if you run your growing cloud service deployments unchecked.

Here are tools and best practices you can deploy to make sure you are not spending too much on your cloud services.

The State of Analytics in IT Operations

What are you really spending?

Public cloud services definitely help slash capital expenditures. Still, it can be difficult to know what you're actually spending on these services, particularly as your consumption proliferates throughout your enterprise.

Gartner reports that the IaaS market grew by 37% in 2017, and it predicts that SaaS growth will double between 2016 and 2020. Facing such high volumes of usage, enterprise IT professionals should take an inventory of what they're using and evaluate whether they've matched the right workloads to the right services to get the best economics.

In other words, you might be saving money with cloud services, but it's entirely possible that you’re still overpaying—or wasting money on services that aren't used or that are underused. Industry studies estimate that 30% to 35% of enterprise dollars spent on public cloud services are wasted

But determining how well your business is doing at managing its cloud expenses is easier said than done. Several challenges surrounding cloud management and pricing have made keeping up with enterprise cloud deployments difficult. Here's a look at those, and ways to overcome them.

Lack of visibility

Do you have visibility into all the cloud services you're using, who's using them, for what purpose, and at what cost? It’s critical that you do. However, getting this top-down view into your public cloud environment gets increasingly difficult as the number and types of cloud services you're using expand in size and complexity.

Cloud providers tell you how much you're spending each month for their services, but they don't break it down by cost center, application, or workload. That makes it hard to determine if workloads have the capacity and configuration that will give you the best possible return.

The specter of shadow IT

The IT department is often not the only group in your organization procuring cloud services. Departments throughout the business may be subscribing to services that you know nothing about. Cloud providers let anyone with a credit card create a cloud server in seconds, contributing to the disaggregation of IT and, in many organizations, causing cloud usage to go largely unchecked.

The impact is a bit like turning a teenager loose with your credit card: Your end-of-the-month bill is always a surprise. Adding insult to injury, shadow IT departments can use the cloud service they provision for a while, and then abandon it, leaving the subscription active but idle and continuing to rack up charges.

In fact, research conducted last year by Nutanix indicated that 5% to 10% of provisioned cloud services are unused in most enterprises, while 10% to 20% are underutilized.

Cloud cost governance and management tools can help provide visibility into under-used and unused cloud resources, and optimize their cloud usage efficiency on a daily basis.

Complex and ever-changing cloud pricing

Knowing your true costs is also difficult because cloud service providers tend to have complex pricing models. They compute their prices based on many usage dimensions and update those prices frequently.

As a result, what might have been the most economical pricing tier for a particular workload yesterday might not be the best fit for it today. Staying on top of these fluctuating prices, plans, and tiers requires a level of real-time diligence that most IT staff are unaccustomed to providing.

Rightsize and automate

Companies accustomed to signing long-term service provider agreements for volume discounts tend to buy lengthy cloud service plans and don't focus on matching specific applications to the best service or service tier.

Enterprises need to get out of this mindset and into a dynamic mode to capitalize on the best price for each workload. In other words, you need to evolve your best practices and processes to mirror the real-time, continually changing nature of cloud workloads and provider pricing.

One way to do this is by taking advantage of automation. Cloud management platforms exist that are independent of the SaaS, IaaS, and PaaS providers whose services you use. Such tools provide visibility across your entire hybrid cloud environment—private clouds and multiple providers' public cloud services—so you can see exactly what's running where.

Using analytics, these will compare pricing across services and tiers at any given time, so you can adjust your workload delegation accordingly. And they can identify any contracted service or resource that hasn't been used for a length of time that you specify. This can help you weed out unused services (or under-used services) that may have been procured outside of the IT department but that are still being paid for.

As for shadow IT, the best way to attack the problem is to let engineering teams use cloud services as needed but centralize the cloud governance through the IT team for all cloud service accounts across different teams in the company.

[ Conference: ITOM Summit 2019 (Feb 5-7, Phoenix, AZ) ]

Figure out where to run workloads

For up-front planning purposes, here are some high-level rules of thumb for where to run different types of workloads.

Static workloads

These are very predictable in nature, and run best—meaning at the lowest cost—in your private cloud. It's fairly easy to forecast the volume of resources that static workloads will consume, and you can plan ahead to incrementally add more resources to your established infrastructure as needed to accommodate them.

Dynamic workloads

These are unpredictable in nature and run best in a public cloud service that you can turn on and off quickly. Depending on sales and seasons, for example, retailers might have unpredictable workloads with erratic usage best served by a service that can be dialed up and torn down on demand—and paid for only when used.

Semi-dynamic workloads

These can run either in a private or public cloud. Which option is best for you depends on the robustness and availability of the data center resources you already have. If accommodating new or additional workloads involves significant on-premises build-out, it might be simpler and cheaper to turn to a public cloud service.

In that case you'll need to compare prices across providers and service tiers and periodically revisit those frequently changing prices to adjust your allocations accordingly.

How to optimize for cost in the public cloud

The dynamic nature of your public cloud workloads means that optimizing your cloud costs isn't a set-it-and-forget-it process. Automated cloud management tools can help by providing the automation and analytics needed to stay on top of your deployments and save money on an ongoing basis.

Once you know which applications and workloads are best suited to a public cloud service (those you've identified as least predictable), you can compare costs among a range of public cloud providers and service tiers and move your workloads to the options that best fit your needs and budget.

Your cost comparison should include evaluating the most cost-effective regions and instance sizes for each application and workload. It should also identify the best discount options based on your usage level.

Repeat this exercise frequently to make sure you're up to date with the best service match for your workloads, and monitor your usage so you meet any commitment minimums and take advantage of existing discounts.

In addition:

  • Identify and deactivate idle or underutilized cloud services that you discover.
  • Run developer resources only when needed. You might try shutting them down at night, for example.
  • Begin to forecast usage. Do this by building a budget for each department or application you move to the cloud. By consulting reports that have been generated by your cloud management platform, you can see individual and consolidated data and begin to build forecasts.
  • Check in regularly with each department or individual involved in cloud spending. Show them reports that indicate where they might be approaching budget limits.
  • Learn the pricing structure for the cloud providers you use, so you can easily identify the right cloud service options from them that will meet your needs while giving you the optimum price.

Make no assumptions

If you've embarked on a cloud-first competitive strategy, you need to be able to control cloud usage and spending. If you merely assume you're saving money compared to building and managing infrastructure yourself, you risk your cloud-computing expenses eventually getting out of hand.

That's why it pays to keep in mind that the model for monitoring and controlling costs has changed dramatically in modern IT departments. Where IT once involved one-time investments with static maintenance agreements, cloud services demand constant oversight and control over hundreds or thousands of instances. Again, this is an unprecedented level of attention never before required of IT.

Fortunately, automated tools have now become available to allow you easily monitor cloud usage in real time and rejigger your deployments to optimize your workflows regularly—without requiring massive investments in human resources.