Project

General

Profile

Actions

Fix #9601

open

erasure-code: ErasureCode::encode overhead is too high

Added by Loïc Dachary over 9 years ago. Updated over 4 years ago.

Status:
New
Priority:
Normal
Assignee:
-
Category:
-
Target version:
-
% Done:

0%

Source:
other
Tags:
Backport:
Reviewed:
Affected Versions:
ceph-qa-suite:
Pull request ID:
Crash signature (v1):
Crash signature (v2):

Description

When encoding 4KB buffers it is ~15% of the total CPU being used although it is only preparing the buffers.
benchmark -O3


Files

kcachegrind.png (282 KB) kcachegrind.png benchmark -O3 Loïc Dachary, 09/26/2014 03:39 AM
jannau-bench.png (310 KB) jannau-bench.png https://github.com/ceph/ceph/pull/2558 bench Loïc Dachary, 09/26/2014 05:17 AM
Actions #1

Updated by Loïc Dachary over 9 years ago

  • Description updated (diff)
Actions #2

Updated by Loïc Dachary over 9 years ago

  • Description updated (diff)
Actions #3

Updated by Loïc Dachary over 9 years ago

  • File deleted (Screenshot from 2014-09-26 12_32_18.png)
Actions #4

Updated by Loïc Dachary over 9 years ago

Actions #5

Updated by Loïc Dachary over 9 years ago

Applying https://github.com/ceph/ceph/pull/2558 and benchmarking again

Actions #6

Updated by Loïc Dachary over 9 years ago

  • Description updated (diff)
Actions #7

Updated by Loïc Dachary over 9 years ago

The overhead has shifted but looks globaly the same with https://github.com/ceph/ceph/pull/2558
https://github.com/ceph/ceph/pull/2558 bench

Actions #8

Updated by Janne Grunau over 9 years ago

The handling got more complicated due to the updated padding handling.

It's a bit little faster. jerasure_matrix_encode's part increases from 76.5% to 77.5%. The effect on the speed reported by ceph_erasure_code_benchmark should be a little larger.

Actions #9

Updated by Patrick Donnelly over 4 years ago

  • Status changed from 12 to New
Actions

Also available in: Atom PDF