book

Systems Performance, 2nd Edition

by Brendan Gregg

December 2020

Intermediate to advanced

928 pages

28h 18m

English

Pearson

Read now

Unlock full access

Includes

Has Sandbox

About This EditionAbout This Book

1.1 Systems Performance1.2 Roles1.3 Activities1.4 Perspectives1.5 Performance Is Challenging1.6 Latency1.7 Observability1.8 Experimentation1.9 Cloud Computing1.10 Methodologies1.11 Case Studies1.12 References
2.1 Terminology2.2 Models2.3 Concepts2.4 Perspectives2.5 Methodology2.6 Modeling2.7 Capacity Planning2.8 Statistics2.9 Monitoring2.10 Visualizations2.11 Exercises2.12 References
3.1 Terminology3.2 Background3.3 Kernels3.4 Linux3.5 Other Topics3.6 Kernel Comparisons3.7 Exercises3.8 References
4.1 Tool Coverage4.2 Tool Types4.3 Observability Sources4.4 sar4.5 Tracing Tools4.6 Observing Observability4.7 Exercises4.8 References
5.1 Application Basics5.2 Application Performance Techniques5.3 Programming Languages5.4 Methodology5.5 Observability Tools5.6 Gotchas5.7 Exercises5.8 References
6.1 Terminology6.2 Models6.3 Concepts6.4 Architecture6.5 Methodology6.6 Observability Tools6.7 Visualizations6.8 Experimentation6.9 Tuning6.10 Exercises6.11 References
7.1 Terminology7.2 Concepts7.3 Architecture7.4 Methodology7.5 Observability Tools7.6 Tuning7.7 Exercises7.8 References
8.1 Terminology8.2 Models8.3 Concepts8.4 Architecture8.5 Methodology8.6 Observability Tools8.7 Experimentation8.8 Tuning8.9 Exercises8.10 References
9.1 Terminology9.2 Models9.3 Concepts9.4 Architecture9.5 Methodology9.6 Observability Tools9.7 Visualizations9.8 Experimentation9.9 Tuning9.10 Exercises9.11 References
10.1 Terminology10.2 Models10.3 Concepts10.4 Architecture10.5 Methodology10.6 Observability Tools10.7 Experimentation10.8 Tuning10.9 Exercises10.10 References
11.1 Background11.2 Hardware Virtualization11.3 OS Virtualization11.4 Lightweight Virtualization11.5 Other Types11.6 Comparisons11.7 Exercises11.8 References
12.1 Background12.2 Benchmarking Types12.3 Methodology12.4 Benchmark Questions12.5 Exercises12.6 References
13.1 Subcommands Overview13.2 One-Liners13.3 perf Events13.4 Hardware Events13.5 Software Events13.6 Tracepoint Events13.7 Probe Events13.8 perf stat13.9 perf record13.10 perf report13.11 perf script13.12 perf trace13.13 Other Commands13.14 perf Documentation13.15 References
14.1 Capabilities Overview14.2 tracefs (/sys)14.3 Ftrace Function Profiler14.4 Ftrace Function Tracing14.5 Tracepoints14.6 kprobes14.7 uprobes14.8 Ftrace function_graph14.9 Ftrace hwlat14.10 Ftrace Hist Triggers14.11 trace-cmd14.12 perf ftrace14.13 perf-tools14.14 Ftrace Documentation14.15 References
15.1 BCC15.2 bpftrace15.3 References
16.1 An Unexplained Win16.2 Additional Information16.3 References
Physical ResourcesSoftware ResourcesA.1 References
CPUsMemoryFile SystemsDisksNetworking
Chapter 2—MethodologyChapter 3—Operating SystemsChapter 6—CPUsChapter 7—MemoryChapter 8—File SystemsChapter 9—DisksChapter 11—Cloud Computing
E.1 References

Content preview from Systems Performance, 2nd Edition

Chapter 9

Disks

Disk I/O can cause significant application latency, and is therefore an important target of systems performance analysis. Under high load, disks become a bottleneck, leaving CPUs idle as the system waits for disk I/O to complete. Identifying and eliminating these bottlenecks can improve performance and application throughput by orders of magnitude.

The term disks refers to the primary storage devices of the system. They include flash-memory-based solid-state disks (SSDs) and magnetic rotating disks. SSDs were introduced primarily to improve disk I/O performance, which they do. However, demands for capacity, I/O rates, and throughput are also increasing, and flash memory devices are not immune to performance issues.

The learning ...