As somebody that has pushed some proprietary services to 4.5M RPS at low-latency (~850-900us p99 in the same dc) I'm a bit skeptical about the "5M+ GET/s" claim.
I'd like to see how the benchmark was performed, on what hardware, with what client-side latencies (ok p50, but what about p99/p99.9?).
A few tables with no methodological details sound a bit fishy.
I'd like to see how the benchmark was performed, on what hardware, with what client-side latencies (ok p50, but what about p99/p99.9?).
A few tables with no methodological details sound a bit fishy.