Remote socket memory access latency
A memory-bandwidth-heavy in-memory analytics service was migrated to a new dual-socket bare-metal host with more total cores and RAM, but per-query latency got ~25% worse and is now erratic, with p99 swinging run-to-run. CPU is not saturated. perf shows high memory-access latency and a large fraction of remote-socket memory accesses; the old single-socket box had none. The process is not pinned and the OS freely migrates threads across both sockets. How do you triage and mitigate?
What a strong answer looks like
Stop the bleeding first (mitigate), then form hypotheses from real signals. Separate root cause from symptom, communicate status as you go, and close with what prevents a repeat.
0:00 of about 40 min
Which questions mattered is sealed until you submit. Telling you now would just be handing over the edge cases.
Run or narrate your approach, then ask the coach.