Opsani leverages neural networks and machine learning to optimize cloud infrastructure and slash cloud bills. Pretty awesome. So awesome that people often ask us how we do it.
One of our talented engineers, Eric Kalosa-Kenyon, has written a technical whitepaper on black box optimization benchmarking that answers this question in depth.
In the whitepaper, Eric describes our approach to verifying optimization performance in light of two major constraints: opaque application internals and costly experiments. Erik describes how he supports Opsani’s benchmarking toolset and how, using a variety of data-driven and de novo synthetic applications, he benchmarks the Opsani optimization engine in challenging scenarios. These scenarios test the engine’s ability to effectively optimize applications whose performance response curves may be noisy and nonlinear. The soobench R package provides synthetic response curves with many of the challenging characteristics we require to refine our software.
Benchmarking our AI is a meticulous process. If you are interested in learning about how Opsani optimizes cloud infrastructure and application parameters, this whitepaper outlining our approach to black box optimization is for you.