I’ve been running some initial performance tests of influx running on an 8-core machine with NVME drives (NVME = very high performance SSD drives). The test was a write-only test writing 20 series in batches of 100 points, each point with ~10 integer fields). I compared this against running using EBS (elastic block store). I was expecting influx on NVME to completely blow away influx on EBS. But I only saw a <20% improvement. It looked like influx was heavily CPU bound (all 8 cores completely pegged by influx). So I used perf to get stack-samples and generate a flame graph: https://drive.google.com/file/d/1tlj5x0-_mmp_DS2qKMXtc9kznbKk6kcm/view (download and open in chrome).
Is the performance profile shown by the flame graph expected?