Optimizing Linux Servers
summary
Advice focuses on concrete knobs and measurements: use custom partitioning and minimal install, detect bottlenecks with top, iostat, vmstat, netstat and benchmark storage with bonnie++ or dbench, change the IO scheduler on the fly via /sys/block/sda/queue/scheduler (echo NOOP or deadline; CFQ is default), enable hugepages for Oracle, run numactl daemon for NUMA, drop caches with echo 3 > /proc/sys/vm/drop_caches, reduce swappiness from default 60, and increase TCP buffer limits and choose congestion control (cubic/htcp) when testing real workloads.
AI-written summary · about
chapters (31)
- 0:00Optimizing Linux Servers
- 1:48What is optimization?
- 2:58Server optimization
- 4:18Start here ... with installation
- 6:37Performance monitoring
- 8:36Context Switch
- 9:11Run queue
- 11:10CPU Utilization
- 12:15LOAD average
- 14:09Disk I/O optimization
- 17:05System scheduler ... how to change
- 17:48More about System scheduler
- 19:251/O optimization
- 22:46Disk optimization
- 23:50Network Tuning
- 25:39TCP Tuning (10GB NIC)
- 26:32TCP Congestion Avoidance Algorithms
- 30:02More about Memory optimization ...
- 31:05NUMAD
- 32:17Huge Pages
- 33:27Flushing Caches
- 34:20Swappiness
- 36:13System Monitoring Tools
- 38:11Kernel Tuning
- 41:25Samba tuning
- 42:08Database optimization
- 42:39OpenLDAP Tuning
- 43:10Apache Tuning
- 43:31Slow websites
- 44:32Benchmark Tools
- 45:21Identify and solve performace issue
related talks
55mPerformance tuning of Red Hat Enterprise Linux
47mPractical Performance Analysis
15mOptimizing Application Performance on Kubernetes
40mAchieving the Ultimate Performance with KVM
37mScaling LAMP doesn't have to suck
40mInspecting and Optimizing Memory Usage in Linux
42mWhen the OS gets in the way
50mDesigning for Performance