PostgreSQL at 20TB and Beyond (Slides)

slideshare.net

4 points·einhverfr··4 comments

I think the biggest fundamental takeaway here is that PostgreSQL works fairly well as a big data platform for certain kinds of data.

20TB per server x 20 servers for raw storage, 32TB of storage sharded for analytics. trillions of data points per year. And we make it work.

They run on "stripped down Gentoo" Linux. They have "about 200 employees worldwide". They also get 50% faster throughput than stock Linux systems.

What's all this "no business justification" for having your own quasi-distro? "Business justification" is just intellectual lazines on the part of The Business itself. Or maybe sub rosa "emoluments", like a few rounds of Bikini Golf in the Bahamas for the VPs.

One fundamental problem here is that business justification depends on business direction and strategy. What is your exit strategy? Do you care about being profitable? How important is autoscaling? How do you deal with a hardware outage?

These are not free tradeoffs. In all cases you solve some problems at expense of either making other problems harder to solve.

So I don't know about the no business justification argument. That is so contextual.

I do know we hire Gentoo packagers. We hire people who can code C. We hire PostgreSQL backend hackers. We hire the best and most senior people we can.

Thank you for an intelligent answer, one that accounts for nuance.

Reply on news.ycombinator.com