How to Scale to 3 Billion Daily API Requests

In 2019, Kevel officially crossed the 2B+ daily API requests mark, a feat few API platforms have reached. Not only that, but we've hit this volume even with SLAs that specify 99.9999% uptime and response times below 100ms.

(Update - as of Jan 2020, Kevel crossed 3B+ daily API requests.)

Of course, it hasn't been a seamless process. Who would have thought the tech stack needed to handle 1M daily requests wouldn't work for 1B?

Needless to say, we've seen our ad serving infrastructure change greatly over the past decade, our wake littered with tools useful at the time but incompatible with scale (see: MongoDB, ASP.net, and many many others).

I recently gave a talk at GlueCon about this evolution, and I wanted to share the presentation below. If you're an engineer or product manager looking to scale your ad server (or any API platform), I recommend looking at the pros and cons of our current and past set-ups.

We're pretty happy with where we are now (as the presentation shows, it's substantially more complicated than where we started), but we expect this won't be our stack at 10 billion daily API requests.

Interested in Kevel?

We are more than happy to set up a 15-min consultation to walk through your needs and discuss the best path forward.