Plone Performance, Profiling, Power-Consumption
1 like226 views
The document discusses performance profiling tools for Python applications, highlighting tools such as py-spy and repoze.profile for analyzing performance issues. It details improvements made in Plone versions 5.2.1 to 5.2.3, including optimizations in caching and memory consumption. Future considerations include optimizing plone.restapi and reducing unnecessary calls to plone.registry.
Plone Performance, Profiling, Power-Consumption
- 8. How to find the performance problem? Tools
- 9. Tools mainly used ● py-spy runtime profiler for Python programs ● repoze.profile Python profiler integration with WSGI ● dis Disassembler for Python bytecode
- 10. py-spy ● Overall mix of the whole application performance ● Unix „top“ like output of running python program ● recording mode: for later analyzing ● Able to attach to live process w/o slowing down Plone
- 11. repoze.profile ● Python profiler integration as WSGI middleware ● Slows down the application ● Ability to profile a single request and analyse its call stack by count/ single call time, accumulated time/ callee/ caller
- 12. Plone 5.2.0 to 5.2.3 plone.dexterity ● avoid early providedBy calls ● __getattr__ early exit on common attrs ● caches optimized: assignable, schema cache ● removal of superfluos threadlocal sync
- 13. Plone 5.2.1 to 5.2.2 zope.interface ● ~1.25x -1.5x __hash__ override removals on Implementer leftover ● ~1.25x __hash__ improvement InterfaceClass ● ~5x lookup/lookupAll/subscriptions ● memory consumption reduced (jmadden)
- 14. DEMO TIME
- 15. Future Todo‘s ● plone.restapi has optimization potential → all navigation related ● plone.registry is called too often ● Use „python:“ expressions in all page templates ● More introspection
- 16. Advice: Start introspecting your performance! Virtual hugs and stay healthy! (Q&A at Jitsi face2face as announced)