-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/70469/
-----------------------------------------------------------
(Updated April 16, 2019, 9:39 a.m.)
Review request for mesos and Benjamin Mahler.
Changes
-------
Updated benchmark result.
Bugs: MESOS-9723
https://issues.apache.org/jira/browse/MESOS-9723
Repository: mesos
Description
-------
This patch optimizes the permutation step of the weighted
shuffling function used by the random sorter.
The permutation step now has constant memory cost and
O(n) time complexity.
This is based on the algorithm described in:
https://s.apache.org/o7K0
Also added a unit test.
Diffs
-----
src/master/allocator/sorter/random/utils.hpp
1329359040082ed128d8dbc45381aa3e529224e4
src/tests/sorter_tests.cpp 9d52a80eafb6f955386a6575875daacf5d4b4e9e
Diff: https://reviews.apache.org/r/70469/diff/1/
Testing (updated)
-------
make check
Ran a modified sorter benchmark:
Before: 100000 runs of sort of 2000 clients took 9.113634ms
After: 100000 runs of sort of 2000 clients took 8.802126ms
It appears to have minimal performance improvement.
Thanks,
Meng Zhu
