SSD and HDD Performance Testing
summary
Open-source TKperf runs SNIA PTS v1.0 workloads using fio, Python plotting and hdparm for secure erase, produces PDF reports, and enforces direct I/O and preconditioning to steady state (<=20% variation over five runs). Measured peaks include ~1.1M IOPS on a PCIe accelerator, ~60k IOPS and ~500 MB/s on enterprise SSDs, ~330 IOPS on HDDs; saturation tests write 4× device size and consumer SSDs drop after ~25 rounds while enterprise models stay stable. Uses refill_buffers to measure incompressible performance.
AI-written summary · about
chapters (13)
- 0:00SSD and HDD testing intro
- 1:04TK perf development
- 2:17TK perf tools and workings
- 4:31Test workflow and steady state
- 7:27IOPS results and PCIe cards
- 13:59SSD vs HDD IOPS
- 16:33Throughput and HDD drop
- 18:06Saturation test: enterprise vs consumer SSD
- 20:21Latency test: consumer vs enterprise SSD
- 21:28Future plans and open-source
- 22:34Performance testing pitfalls
- 23:38Dedicated server setup
- 25:25Closing remarks and Q&A
related talks
40mBenchmarking: Repeatable & Comparable
04mArtur on SSD's
43mComparing Performance of NVMe Hard Drives in KVM, Baremetal, and Docker
15mYears of Wear in Minutes: Benchmarking Aged Filesystems
34mEvaluating the Performance and Power Efficiency of Ceph Storage Configurations
04mLightning Talk - Benchmarking Persistent Data Volumes
38mScaling IO-bound application in the cloud
40mPerformance Testing Done Right