The Hardware behind the Results of Framework Benchmarks Round 10

By Jeffrey Papen, CEO and Founder, Peak Hosting

At Peak Hosting, we're big fans of TechEmpower's Framework Benchmarks, an open source project the company has been coordinating since early 2013. Covering a wide variety of web application frameworks, this project gives developers useful data that can help them find the framework that will provide the performance and features they need for their application.

TechEmpower's benchmarking now includes six test types, more than 120 frameworks, 290 test permutations, and results that include latency and framework overhead.

Hardware comes into play when performance is important. And TechEmpower will tell you performance is always important. The best results were derived from a real-world environment running physical hardware. As a managed hosting provider, we were able to provide the project with the same types of machines that our customers use to run their production environments.

We first contributed to TechEmpower's Framework Benchmarks in Round 9 when we set up for the project five dedicated Dell R720 dual-Xeon E5 servers with 10 Gigabit Ethernet running in our data centers. High-end hardware directly correlates to high performance and the results from Round 9 to Round 10 bear this out. According to TechEmpower's Round 10 blog post:

Competition for the top position in the JSON-serialization test within the Peak Hosting environment has heated up so much that Round 10 sees a more than 100% increase in the top performance versus Round 9 (2.2M versus 1.05M). A year ago, TechEmpower showed that one million HTTP responses per second without load balancing was easy. We're delighted that 1M is already old news.

So what does hardware have to do with this impressive round-over-round improvement? We didn’t change the hardware between Rounds 9 and 10. What did change was that between rounds, test implementation contributors realized they had hardware available to them with 40 hyperthreading cores, and they were able to optimize their code to take advantage of that performance and capacity. A bit of tweaking for high-end hardware was all that was needed to utterly smash the previous round's leaderboard.

We're pleased that we're able to provide TechEmpower and the open source community with this hardware environment—and that it's the same type of hardware our customers use every day in production environments, making the results as valuable as possible. And we will eagerly await the results of Round 11 where we anticipate more significant performance leaps!