Performance tuninng for Red Hat Enterprise Linux.
summary
RHEL tuning uses tuned profiles via tuned to apply kernel settings at boot and runtime, and shows concrete gains: about 5% CPU, ~20% Java, and 15–50% disk/DB improvements versus RHEL7; small TCP packets improve ~25–30%. Best wins come from NUMA pinning and swap_eNOS/zone_reclaim tuning, using hugepages (static or dynamic 1GiB) and disabling transparent hugepages for latency‑sensitive DBs, plus using drop_caches for repeatable benchmarks. RHEL8 adds 5‑level page tables, NVDIMM support and Cgroups v2.
AI-written summary · about
chapters (15)
- 0:00Intro
- 4:03RHEL tuned parameters that effect performance
- 8:13RHEL 8 vs RHEL 7 Workload Performance Gains
- 9:27RHEL 8 Performance Coverage
- 14:10RHEL 8 - Database tuning tips
- 17:42RHEL 8 Performance Open Source DBS
- 23:12RHEL 8 Performance Legacy DB
- 25:47RHEL 8 Performance DB Performance
- 26:09RHEL 8 Performance Summary
- 27:09Typical Four-Node NUMA System
- 28:17Tools to display CPU and Memory (NUMA)
- 33:50Interaction between VM Tunables and NUMA
- 38:572MB standard and 1GB Hugepages
- 41:48Transparent Hugepages
- 45:57dirty_ratio and dirty_background_ratio
related talks
55mPerformance tuning of Red Hat Enterprise Linux
25mScheduling and Load Balancing in RHEL8 and Beyond
51mLinux Systems Performance - Architecture
51mTuning RT Kernel to Improve Schedule Performance with Intel Platform- 51mTuning RT Kernel to Improve Schedule Performance with Intel Platform
36mTuning Memory and the Network Stack in Linux
39mTuned - helper for system tuning
31mScaling & Optimizing Performance on AMD EPYC Systems