It would be nice to extend our nightly benchmarks to incorporate update storms which I have replicated with modifications to NRTPerfTest (PR) that simulate update storms with war & peace times. The modified test alternates between "war time" and "peace time" multiple times, creating a more rigorous benchmark that reflects some real-world usage patterns.
I've generated example graphs showing the performance characteristics during these alternating high/low update periods:
It would be nice to extend our nightly benchmarks to incorporate update storms which I have replicated with modifications to
NRTPerfTest(PR) that simulate update storms with war & peace times. The modified test alternates between "war time" and "peace time" multiple times, creating a more rigorous benchmark that reflects some real-world usage patterns.I've generated example graphs showing the performance characteristics during these alternating high/low update periods:
