Performance/MemShrink/Meetings/2011-12-13

Statistical analysis of memory-related telemetry data

•  :Joy from Metrics looked at memory_resident across 7, 8, 9. Other variables such as OS, CPU count, memory size, uptime, etc. How much of the variation can be attributed to these various factors? Analysis of variance (ANOVA). Need some kind of human review of assumptions, etc. to make sure the numbers are reasonable. A combination of domain knowledge and statistical knowledge. njn expressed great enthusiasm towards helping him. Joy is going to do some more analysis and get back to us. He's based in SF.

B2G MemShrink

• responding better to memory pressure events
  • ways to drop memory without using much, esp. JS
  • style, layout data
  • iframes
• slimming down child processes
  • better sharing between content processes
  • currently only share code?
  • lighter weight process would allow resetting more often to fix fragmentation
• Taras has filed a bug about putting the network cache in its own process, so it can be easily dumped.
• njn will file bugs for some of these ideas.

Upcoming meetings

• 12/20: Meeting, no jst.
• 12/27: No meeting.
• 1/3: Meeting!