~/talk

GC Tuning Confessions Of A Performance Engineer

Talkat Strange Loop 2015 · 2015-09-24..09-26· [databases]
slides:none on file

Claims concrete fixes: use GC logs plus VisualVM and simple parsing scripts to plot occupancy and scaling, and expect to find worst‑case pauses (one case shows a 20,699s full GC). Compares Parallel Old vs CMS vs G1 with actionable tunables: marking threshold 45% and G1 Reserve percent 10% (capped 50%), region sizes 1–32MB, PA time goal examples like 5s. Describes evacuation failures, humongous allocations, and practical knobs that do and do not avoid full GC.

AI-written summary · about

  1. 0:00Intro
  2. 1:23About Today's Talk
  3. 2:36Performance Requirements
  4. 3:17Response Time Metrics
  5. 4:42Performance Analysis
  6. 5:58GC Performance Engineering
  7. 7:13Fun Facts!
  8. 7:28Ideal GC?
  9. 7:39The Reality!
  10. 8:55The Tradeoff
  11. 9:19The Throughput Maximizer
  12. 11:00Mr. Latency Sensitive
  13. 12:18The Serial Collector
  14. 13:42The CMS Collector
  15. 15:14Java Heap
  16. 16:09Contiguous Old Generation
  17. 17:00Concurrent Mode Failures
  18. 17:43The Garbage First Collector - Regionalized Heap
  19. 21:46Avoiding Evacuation Failures
  20. 24:48G1 Heap Waste Percentage
  21. 26:11The Garbage First Collector - Humongous Objects
  22. 29:11Generation Sizing
  23. 30:11Scaling: GC Times vs Sys Times
  24. 30:35Reference Processing Times
  25. 31:13GC Overhead vs Elapsed Time
  26. 33:33Old Generation & Collections
  27. 36:29GC Tuneables - The Throughput Collector
  28. 38:17GC Tuneables - The G1 Collector
  29. 42:23Further Reading
~/search
type to search 6,249 conferences, 138,987 speakers and 185,172 talks
~/help
cmdctrl k
search every conference, speaker and series
/
focus the filter search
h
home
c
conferences
s
speakers
e
series
t
light / dark — the status bar button does the same
?
this help
esc
close this, then clear every filter — or just the search, while you are typing in it

Ignored while you are typing. Pages without a filter rail have no search to focus.