Installation time is important as shorter installation allows everyone to safe resources and their time. It also can make better impression when starting with openSUSE and SUSE. And last but not least as we more frequently use openQA, so we do hundreds of installation every week, it even smaller improvement can in the end help SUSE to safe resources and be more green.
Result of Project
installation on testing VM speed up for minimal installation by 11 seconds ~ 7% speed up and for gnome installation by 17 seconds ~ 1.3% speed up
bunch of pull requests created , , , , , and to speed up installation
ruby native profiler added to YaST. Can by activated by Y2PROFILER=1
fixed broken ruby termination in yast2-ruby-bindings as it is discovered by added profiller
Looking for hackers with the skills:
Nothing? Add some keywords!
This project is part of:
Hack Week 14
Activity
Comments
-
over 8 years ago by jreidinger | Reply
First result: speed up from 58 seconds to 45 seconds by bootloader writting optimization in bootloader finish step. related pull requests: https://github.com/yast/yast-installation/pull/398 https://github.com/yast/yast-kdump/pull/61
-
over 8 years ago by jreidinger | Reply
next step is profiling installer run to recognize where to focus on gist
-
over 8 years ago by jreidinger | Reply
Another generic optimization in logging to use more specialized caller call improve minimal autoyast installation by 5 seconds, which is around 3% of whole installation time.
-
over 8 years ago by jreidinger | Reply
Another step. Trying to optimize DeepCopy and found that new version is actually slower. For benchmark see https://gist.githubusercontent.com/jreidinger/917d5681c87b7645bcf6/raw/8dc32e1091f23db1ef47b392c73048b58daf0edc/deepcopyprofile.rb
-
over 8 years ago by jreidinger | Reply
https://gist.githubusercontent.com/jreidinger/917d5681c87b7645bcf6/raw/8dc32e1091f23db1ef47b392c73048b58daf0edc/deepcopyprofile.rb
-
over 8 years ago by jreidinger | Reply
Removing autoyast installation sleep saves 4 second in auto installation pull request
-
over 8 years ago by jreidinger | Reply
Another set of benchmarks done to try to optimize some often used methods, but no success ( it usually speed up only in subset of cases ). But with profiling found that slide show often counted total size, so optimized in this pull request hopefully it should save another seconds ( measuring in progress ).
Similar Projects
This project is one of its kind!