See how TSO Logic
simplifies IT transformation.

Our software platforms remove the guesswork so you can now plan with confidence.

See how TSO Logic
simplifies IT transformation.

Our software platforms remove the guesswork so you can now plan with confidence.

Research: AWS Cloud Costs Keep Declining

Posted in: Blog

The Numbers Don’t Lie: AWS Cloud Costs Keep Getting Lower.

 Waiting for the perfect time to start your migration to Amazon Web Services? If cost savings is part of your decision criteria, there’s good news: it’s never been less expensive to move to the cloud. New research from TSO Logic shows that the prices for AWS keep getting lower, and that those declines are pretty durable. The more of your resources you move to the cloud, and the longer you keep them there, the more value you can expect to get for your money.

Don’t believe it? Let’s run through the data.

AWS Costs Show Steady Declines

If you’ve been paying attention to the cloud marketplace, you probably have a general sense that prices are declining. AWS alone has cut prices more than 60 times over the last few years. But it can be easy to dismiss advertised price drops as marketing hype. We decided to find out what those ostensibly lower prices mean when applied to real-world workloads that enterprises are running. We looked at actual compute and storage provisioning, software license types and utilization levels from an anonymized data set of entperiese servers. We then ran those findings through our predictive analytics engine to compare against current and historical Amazon cloud pricing catalogs going back several years.

The results: the workloads these clients are running in 2017 cost 73% less than it would have cost to run them in 2014. The table below has the details. It illustrates the cost of running 100,000 workloads on AWS in 2014 with costs of running like-for-like workloads—those with the same compute and usage patterns—over the following three years. (We skipped cloud server products introduced later, which couldn’t be matched to identical instances from previous years.) The analysis also calculates cost savings when using like-for-like payment models, as well as the added benefit of using Reserved Instances, which were not available until 2016. The results speak for themselves.


Since 2014, the cost of running like-for-like enterprise workloads on Amazon’s cloud has dropped by at least 57% when comparing the same payment types (that is, when comparing On-Demand pricing for both current and past instances). When the same workload instances are converted from On-Demand to Reserved Instance pricing, which became available in 2016, the savings swell to 73%. When looking at a more recent period, 2015-2017, pricing declines were still significant. Since 2015, the cost of purchasing the same On-Demand instance has declined by 14%. When converting to Reserved Instance pricing instead of On-Demand, costs fell by 46%.

Flexible Core Count Options Push Even More Savings

Clearly, AWS has been offering customers more cloud for their money every year. But with Amazon’s recently announced “Optimize CPUs” option, AWS customers have a new option to save even more. By thinking carefully about core count, you can now unlock software license savings that potentially dwarf those infrastructure savings.

The new feature gives you greater control of Amazon EC2 instances on two fronts. First, you can specify a custom number of virtual CPUs (vCPUs) when launching new instances to save on vCPU-based licensing costs. Second, you can disable Intel Hyper-Threading Technology (Intel HT Technology) for workloads that perform well with single-threaded CPUs, like certain high-performance computing (HPC) applications.

This change will be a big deal for AWS customers running database workloads like Oracle and SQL Server on EC2 today, who want high memory, storage and I/O bandwidth, but don’t need a high vCPU count, since these workloads are rarely compute-bound. Previously, AWS offered only fixed EC2 core count and memory configurations. If your SQL Server instance required large amounts of RAM, for example, then you had to buy a larger number of cores—even if the workload didn’t actually need them. Those extra cores translated to higher than necessary licensing costs.

With the new option, you can now disable cores on EC2 instances. Buy a 4xlarge instance, for example, which has 16 vCPUs by default, and you can now configure that instance to run with 1 to 14 vCPUs instead. The price of the instance stays the same, but by disabling those cores, you get hugesavings when you bring your own licenses for SQL, Oracle and other applications licensed on a per-core basis.

Custom Core Count Savings in Action

Let’s put some real numbers behind these savings. One TSO Logic customer was using 1,356 licensed cores of SQL Server on-premise and wanted to bring their SQL license to AWS. Before the new Optimize CPUs feature, they would have had to use 2,220 cores (63% more) on AWS to meet their memory needs. Considering that the list price for Microsoft SQL Server is $14,000 per core, that fixed core count requirement would have meant a huge spike in their licensing costs—effectively making cloud a nonstarter for this workload.

Now, that customer can keep the same memory, storage, and bandwidth of the full-size instance they need. But they can run SQL Server in AWS with just 1,272 cores—7% fewer than they’re currently using on-premises. Running in AWS with 84 fewer cores translates to a significant price drop over this customer’s current environment. Even more important, the core count flexibility means that cloud is now a viable option for their SQL Server. They can bring their existing license and start realizing all the other benefits of cloud.

(Note that this analysis respects Microsoft’s policy of requiring a four-core minimum for SQL Server licenses. If this requirement were not enforced, this customer would be able to reduce their core count by another 258 cores—bringing licensing costs down even lower.)

What’s Driving Steadily Declining Cloud Costs?

The combination of steadily declining instance costs and expanded licensing savings (through “bring-your-own-license” and core count reduction) has made cloud more cost-effective every year. But should you expect that trend to continue? All signs point to yes, due to the underlying factors driving ongoing cloud savings. Cloud providers benefit from a perfect storm of better technologies, better efficiencies and better economies of scale—all of which create a virtuous circle for cloud customers of all sizes.

Take the ongoing advances in processing power and efficiency. Unlike on-premise environments, AWS always has access to the latest, greatest server hardware—including working directly with chip manufacturers to create custom templates to do exactly what they need, in the most efficient way possible. That can make a huge difference in your costs. Compared to last year’s AWS C4 instances, for example, the newest C5 instances, using the latest-generation Intel Skylake processors, deliver double the vCPU power and performance, 2.4x the memory and 3x the throughput.

Since server hardware improves every year, it’s a safe bet that, whenever you migrate, you’ll be able to get the Cadillac of server hardware in the cloud for a great price. Or, you’ll be able to get last year’s Cadillac for the price of this year’s economy car. Either of which may be more than you need—an Intel single-core processor from 2018 may be able to handle workloads that required dual-core processors just a couple years ago.

AWS also benefits from huge and growing economies of scale as more customers jump onboard, along with ongoing advances in its management of cloud-scale infrastructure. The upshot: the cards are stacked for AWS to keep getting better and more efficient at what they do, and to keep passing those savings onto customers.

Don’t Be Afraid to Get in the Game

With the ability to define core counts at a granular level, along with the steady year-to-year declines in AWS instance costs, cloud really is more economical than it’s ever been. And if you’ve been considering moving your Microsoft workloads to AWS, but thought it wouldn’t be cost-effective, it’s time to look again. Bottom line: there literally has never been a better time to move to cloud.

Of course, selecting the right AWS options to maximize the savings gets complicated. To capitalize on continuous improvements in processing power and instance size price cuts, you can’t just migrate once and never think about it again. You have to continually identify the best and most cost-effective instance type for job. To maximize your software license savings, you have to understand which of your existing licenses can move—as well as the best instance type and configuration to move to. As you’re looking for the best BYOL options in the cloud, it’s also probably worth your time to analyze your current environment for core optimization as well.

All of this analysis gets very complex, very quickly. None of it can be done with a spreadsheet or back-of-the-napkin calculations. Fortunately, the TSO Logic platform’s advanced algorithms and predictive analytics can crunch all those numbers, so you don’t have to. Cloud or on-premise, we make it easy to make sure you’re always using the best resources at the lowest possible cost.

Contact us today and we will help you quickly determine your most-cost effecitive path to cloud.


Virtual Strategy Magazine Article
Executive Viewpoint: Four Ways Business will Benefit from Falling Cloud Prices

How much more could you save?

Thanks for trying our savings calculator. Enter your info below and we'll send you a detailed report showing the long-term savings you could see using TSO Logic's open platform.