Hacker News Reader

Sweet write up. I'm curious as to how CloudFlare's needs are similar/different from other companies (i.e. Facebook, Google, Twitter etc.). It seems to me that there would already be quite a lot of iteration and therefore experience/knowledge for what the best set-up would be? Though a large portion seems to be with new hardware coming out.

View on HN

I work for a company building servers for Facebook. They really did try to follow the Open Compute (http://www.opencompute.org/) architecture but ... um, it didn't work out particularly well. They turned to us for tweaking, Quanta for server manufacturing, and us again for rack integration and large scale testing.

<edit> We are the OEM (e.g. design and build) for large scale storage arrays for Amazon & Netflix, too, but not compute servers.

To clarify the pronouns, "they" here does not refer to CloudFlare. It may refer to Facebook, but I hope not as it would certainly be a violation of a NDA.

Reply on news.ycombinator.com