The Real Bottleneck at 1 Million RPS
A real 1M req/sec benchmark on AWS. The database broke exactly once. Here's what broke the other four times, and what actually fixed each one.
$ grep -r "#performance" ~/blogs # my words, with labels
I tagged my opinions.
Even my tangents have an index. I made it easy to keep up.
$ ls ~/write/connections
my posts, connectedI have receipts for that.
2 posts
These are my #performance posts. I made the connection; you get the links.
Find out where your latency actually goes
A measurement procedure that tells you which layer of your stack owns the delay, before you optimize the wrong one. Worked through on a system that turned out to be 98% network.