Controlling Cloud Costs: Ways to Reduce Public Cloud Spend

By Cloud and Data Center Transformation
6/15/2020

Recently, CDCT experts Peter Kraatz, Scott Cameron, and John O’Shaughnessy participated in a LinkedIn Live session entitled, “Take Control of Your Cloud Costs,” examining ways organizations can control cloud costs while still achieving maximum value from cloud solutions. 

In addition to implementing strong cloud governance as means to cost optimization, Cameron, O’Shaughnessy, and Kraatz identified three main ways to reduce public cloud spend. 

1.)  Assess your organization’s compute, storage, and egress 

According to Cameron, assessing your organization’s amount of computing is an easy target for lowering public cloud spend. “Really, you've got three categories of cloud costs. You've got compute, storage, and network. When you look at those, the most expensive cost for most customers is compute.” He notes that GPUs and data warehouses, in particular, tend to be two of the largest consumers. 
 
While both have an “extremely high value for specific workloads,” if they’re not set up with the proper controls and governance at the outset and “you let people spin up whatever they want, they’re going to spin up whatever they want, and it’s going to end up coming back to bite you,” warns Cameron. The result? Runaway spending. 

For O’Shaughnessy, storage represents another area for easy overspending. “You've got the same problem we've seen for decades, which is storage that's orphaned. Storage that's unused and the ease of spinning things up in the cloud makes it more likely that that's going to be a bigger growth area.” 

While network egress charges don’t represent the majority of cloud costs, they are worth examining, too. “If you look at network egress charges, they're not terrible unless you're extracting wholesale tons and tons of data from the cloud. If you're doing backup and disaster recovery, you're mostly pushing to the cloud with periodic restores. If you have a wholesale failover and then fail back, and you need to ship a whole bunch of data back on premises, you may have a one-time hit,” says Cameron.

2.)  Use the cloud for ideal uses — like Disaster Recovery (DR) and backup

Reserving public cloud consumption for ideal uses is another great way to prevent unanticipated spending on public cloud platforms. For example, cloud-native applications, seasonal and cyclical businesses, and backup and disaster recovery functions all lend themselves to a favorable cost model for public cloud use. 

“DR is a great use case,” says O’Shaughnessy. “Ideally, it's already well understood. It isn't day-to-day production, but it is built up to be that way. There should be existing runbooks, processes that allow people to move to a different environment. It's tested, so to take that existing DR-type environment and say that's how we're going to dip our toe into the cloud, it's a great option.” 

O’Shaughnessy goes on to explain that moving DR to the public cloud allows organizations to test features in the cloud. “At the end of the day, you can spin things up and then spin things down. You're not paying money for the bulk of your resources once you're done with your DR activities,” he concludes. 

Similarly, cloud-native applications represent “big wins” in cloud cost optimization, says O’Shaughnessy. Applications born in the cloud allow organizations to take advantage of platform as a service as opposed to direct movement to infrastructure as a service. 

Cameron points to businesses with cyclical compute patterns as potential good candidates for the public cloud. This includes applications that are consistent, “but only happen during a certain period of the month or quarter batch processes, applications, and workloads that can scale based on performance over time.” For instance, tax preparers or the IRS experience a huge increase in demand leading up to tax season. Then, after the deadline passes, their system goes mostly unused for the rest of the year, making the public cloud an ideal option. 

“How cool is it to be able to automate the availability of that stuff based on when it's necessary to be running and then not pay for the compute for that when it's not in use?” he asks. 

For organizations wondering how to determine which applications are ideally suited for the public cloud, O’Shaughnessy and Cameron agree that a workload alignment assessment can provide valuable insight. 

3.)  Implement elasticity and other options 

Taking advantage of elasticity is another way to reduce public cloud overspend. Organizations that successfully control public cloud costs are “looking at building new in the cloud, building native, taking advantage of the elasticity that's available so they can scale up and scale down, so they can use resources in the most efficient way possible. That's where they're being successful,” says O’Shaughnessy. 

For Cameron, there’s a whole list of options where cost-control measures can be implemented. “There's a list that I go through with customers that ranges from licensing, the choice of VM sizes…automation of scale with elasticity, implementing governance with role-based access control, making sure there is a foundation of cost management.” He notes that once things are in place, organizations “have the functionality of that cost management” that can be leveraged and allow for consolidating resources, subscriptions, or tenants.

At the end of the day, organizations do have options to minimize their cloud overspend. As O’Shaughnessy says, “There are a lot of different levers to be pulled, to monitor your costs, to control your costs, when you’re moving to the cloud. You can be very efficient, but you just have to look in different areas,” to figure out where to optimize.