The Silent Performance Killers: BIOS and Firmware Updates
summary
Vendors push BIOS and firmware patches that can silently reduce throughput and increase latency unless you run pre- and post-update benchmarks. Uses redfish calls to pull BIOS settings, stores time-series results in a database and shows graphs in an observability platform, comparing IO CPU and memory tests plus app-specific workloads. Treat ~3% threshold as meaningful and watch cumulative drops (often 3–5% and sometimes >10% after CVE patches). Built automated baseline/compare tooling that exposes regressions so vendors can be engaged.
AI-written summary · about
chapters (10)
- 0:00BIOS and firmware intro
- 1:19BIOS and firmware roles
- 3:30Why firmware updates
- 5:58Server performance baselines
- 7:28Vendors and performance testing
- 9:57Performance degradation over time
- 11:42Testing and observability
- 13:31Working with hardware vendors
- 14:56Firmware update impacts
- 16:00Conclusion and Q&A
related talks
20mKeynote: MDS, Fallout, Zombieland & Linux
05mIgnite: Microcode Updates as Protection Against Spectre & Co.
19mKeynote: MDS, Fallout, Zombieland, and Linux
30mSpectre, Meltdown, & Linux
51mLinux Systems Performance - Architecture
06mQ&A
30mHow Can We Effectively Test Transient Execution Mitigations?
51mTuning RT Kernel to Improve Schedule Performance with Intel Platform