Improving Price and Performance with AWS Spot-Based Instances

Prowess Consulting tested whether spot-based instances are a viable cost-saving strategy for running Monte Carlo workloads on Amazon Web Services® (AWS®).
Improving Price and Performance

Prowess Consulting tested on-demand and spot-based Amazon® Elastic Compute Cloud™ (Amazon EC2®) instances using the financial services workload sample with Monte Carlo European options simulation workload. Intel had earlier run Monte Carlo testing on several Amazon Web Services® (AWS®) instance types with regular on-demand subscriptions, but we wanted to go deeper with more testing. In this new study sponsored by Intel, our testing goals were to:

1. Validate the results of the Intel testing with similar tests on more instance types.

2. Expand the test scope from pure performance to price performance—that is, to measure the cost of performing the same work on different instances and processor types.

3. Explore the feasibility and price performance of using spot-based pricing—a lower pricing tier contingent on availability of excess capacity—instead of on-demand pricing.

At the time of our testing, instances based on 3rd Gen Intel® Xeon® Scalable processors emerged as top performers in terms of throughput and speed. Our testing also showed that instances based on 2nd Gen Intel Xeon Scalable processors were price performance champions.¹ While some instances based on AMD® and AWS Graviton® processors were competitive in terms of raw performance and price performance for on-demand instances, spot-based instances for these processors were often not available at the time of testing.

Contact Us

Interested in working with us?

Ready to get started? The Prowess team would love to discuss the business challenges you’re facing and how we can put our experience into action for you.