- Andrew Midgley, Isaiah Rendorio, and Jing Xu
- November 10, 2020
Apptio Cloudability now calculates the savings you realize from your Amazon EC2 Spot Instance usage and allows you to analyze those savings over any period of time. The release of this new reporting capability by Apptio allows customers to understand the precise benefit realized by running Spot instances, with the platform’s advanced allocation features ensuring granular visibility of these savings across delivery teams, helping surface ROI and educate future provisioning decisions.
How Spot Instances can reduce your hourly rates
Amazon EC2 Spot Instances have been an important option for reducing hourly compute rates going all the way back to 2009. The Spot market was introduced by Amazon to provide access to excess EC2 capacity, with customers potentially receiving heavily reduced rates (up to a 90% discount) compared to on-demand usage. While engineers are generally required to build software applications to handle this market’s ephemeral nature – Spot instances can be terminated by AWS at any moment – this option has proven a boon for many types of expensive workloads, especially those involving big data or batch jobs.
Deciding whether to provision an EC2 instance using Spot is not always straightforward and we at Apptio are excited to play a bigger role in helping our customers identify the financial implications of these decisions and simplify ongoing management of Spot instances.
Prices go up, prices go down
Unlike commitment planning instruments (Reserved Instances and Savings Plans), which require little to no engineering effort to receive a financial benefit, Spot instance usage requires deliberate action at the time of provisioning and often a significant investment architecturally to make software applications stateless or fault tolerant. The benefit of reduced pricing needs to outweigh this extra effort.
The non-deterministic nature of spot pricing (your hourly rates will go up and down day-to-day as supply and demand fluctuate and AWS adjusts the Spot prices) makes quantifying this benefit difficult, especially when you are running cloud resources at scale with many variables at play. In fact, while the parameters you choose at the time of provisioning — maximum spot price or instance type, for example — might initially deliver strong returns, fluctuations in the market could quickly lead to there being superior choices. Another consequence of this dynamic pricing is that it complicates your ability to understand underlying usage trends and predict future spending patterns.
Identify savings and inform future spot decisions with detailed analytics
Cloudability is a full cost analytics platform. We have always set out to answer the widest set of questions possible. One aspect of this has been to provide several different cost metrics that aim to represent your cloud bill in ways that make sense to the business. For example, some stakeholders will examine simple cashflow – Cost (Total) – while the accounting team will undoubtedly need an accrual account perspective – Cost (Amortized). A metric that has served a particular purpose, especially for peeling back the discounting impact of commitments, is Cost (List). Using this metric alongside the others has allowed customers to precisely understand their realized savings and how these savings are trending over time – for example, you can compare the Cost (Amortized) metric to the Cost (List) metric to surface the impact of a reservation on a certain bit of cloud usage . This metric has also proven useful for chargeback and planning activities as it is not impacted by the unpredictable nature of discounting. Interestingly, whereas commitment-based discounting brings unpredictably due to variability in where it is applied, Spot instance discounting introduces unpredictably due to variability in market rates.
With this update, we are thrilled to announce that the same level of reporting is now also available for managing Spot usage. The biggest challenge we had to overcome in getting this consistent behavior was obtaining the On Demand rates that would otherwise have been applicable – information that is not present within AWS’ billing data itself. We are now able to retrieve this historical pricing information across the broad spectrum of SKUs from our data lake and calculate the accurate on demand cost that would have been charged for your resources had you not been running them as Spot instances.
With the Cost (List) metric, you can see what the on demand cost would otherwise have been for instances run as Spot, and by comparing this to Cost (Total), identify the exact savings over any time frame. This data is populated for the full history of Spot, so feel free to run this analysis for previous years. The simplest means to visualize such insights is through dashboards, where margins can be tracked, trends identified, teams compared, and analysis done to the resource level. It is even possible to define a business metric for the purpose of publishing the realized savings across Spot and commitment usage.
To learn more, sign up for a Cloudability free trial today!
