A naukri.com initiative
Github
4w
373
Image Credit: Github
Breaking down CPU speed: How utilization impacts performance
- The Performance Engineering team at GitHub conducted experiments to observe the impact of CPU utilization on system performance.
- The team used a tool called “stress” to gradually increase CPU utilization and established a baseline by sending moderate production traffic to the Large Unicorn Collider (LUC) Kubernetes pod, which mirrored the same architecture and configuration as flagship workloads.
- As expected, CPU time increased for all instance types as CPU utilization rose. Each instance type showed unique behavior and varying thresholds where performance began to decline.
- The team observed that lower CPU utilization leads to better performance, while higher utilization leads to a decline in performance.
- Intel’s Turbo Boost Technology also impacted CPU frequency, leading to a decrease as utilization increased, causing overall system performance to decline.
- All nodes also had Hyper-Threading enabled, which allowed a single physical CPU core to operate as two virtual cores.
- To achieve the optimal balance of CPU utilization, the team identified a threshold where CPU utilization is high enough to avoid underutilization but not so high as to significantly impair performance.
- The team derived a mathematical model to identify the threshold by determining what percentage of CPU time degradation is acceptable and plotting the CPU utilization vs. CPU time (latency).
- The team also discovered an issue where certain instances were not achieving their advertised maximum Turbo Boost CPU frequency due to a disabled CPU C-state, which prevented the CPU cores from halting even when they were not in use, and they successfully resolved it through re-enabling C-state.
- These insights enable GitHub to inform resource provisioning strategies and maximize hardware investments.
Read Full Article
22 Likes
For uninterrupted reading, download the app