What Millions of Requests per Second Mean in Terms of Cost and Energy Savings

Over the past few years, the team behind HAProxy has advanced the technology further than ever, setting new world records in terms of performance and scalability. The benchmark achievements have reached such astonishingly large numbers—millions of requests-per-second processed, many parallel threads, 100 Gigabits-per-second throughput, tens-of-thousands of TLS keys per second—that they’ve become difficult to relate back to most people’s everyday experience. So why are these benchmarks important? What do they really mean when it comes to cost cutting and energy savings?

We are at a moment when climate change has focused our attention acutely on finding ways to conserve in every possible way. It is no longer acceptable to run power-hungry machines, both in terms of costs and in public perception. However, applications cannot be redesigned in the blink of an eye. It therefore becomes critical to find solutions in other areas, such as the load balancer tier, and in this talk Willy Tarreau, the developer of the original HAProxy project, will be exploring real-life examples of the software’s ability to do just that.

Willy Tarreau

Lead Developer, HAProxy Technologies

Willy released the first version of HAProxy in 2001, welcomed the first contribution in 2004 and became a Linux kernel maintainer in 2006. His focus has always been on the lower layers where efficiency can still be improved after everything was squeezed at visible layers, and on reliability, probably because he hates revisiting complex code. By better layering an already robust and efficient core, he aims at making the project more accessible to new contributors interested in having fun doing great things without dealing with the apparent complexity of a full-featured load balancer.