Bug 1284872 - Poor performance of small file / create heavy workloads over SMB
Summary: Poor performance of small file / create heavy workloads over SMB
Keywords:
Status: CLOSED CURRENTRELEASE
Alias: None
Product: Red Hat Gluster Storage
Classification: Red Hat Storage
Component: samba
Version: rhgs-3.1
Hardware: All
OS: Linux
high
high
Target Milestone: ---
: ---
Assignee: rhs-smb@redhat.com
QA Contact: Ben Turner
URL:
Whiteboard:
Depends On:
Blocks: RHGS-3.4-GSS-proposed-tracker
TreeView+ depends on / blocked
 
Reported: 2015-11-24 11:24 UTC by Michael Adam
Modified: 2019-10-10 10:33 UTC (History)
11 users (show)

Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Clone Of:
Environment:
Last Closed: 2017-09-21 09:26:57 UTC
Embargoed:


Attachments (Terms of Use)
Perf stats ion table form showing performance degradation (222.85 KB, application/vnd.oasis.opendocument.spreadsheet)
2015-12-17 17:25 UTC, Harold Miller
no flags Details


Links
System ID Private Priority Status Summary Last Updated
Red Hat Knowledge Base (Solution) 2094901 0 None None None 2015-12-17 17:55:05 UTC

Description Michael Adam 2015-11-24 11:24:02 UTC
Description of problem:

Small file workloads (create/meta data heavy) perform poorly with Samba/SMB on top of Gluster compared to NFS or native client. On the other hand, on local file systems, SMB is fast.

More details to follow.

Comment 3 Harold Miller 2015-12-17 17:25:31 UTC
Created attachment 1106780 [details]
Perf stats ion table form showing performance degradation

Comment 5 Harold Miller 2016-01-04 16:41:44 UTC
Ben,

May I get an update on the SMB - performance issue, please?

Is this a RHEL6, RHEL7, or both fix?

Harold

Comment 7 Harold Miller 2016-01-13 17:26:59 UTC
Status Update please.

Comment 8 Poornima G 2016-08-01 11:33:06 UTC
Harold,

We are working on the perf improvement for samba access, here are the patchsets http://review.gluster.org/#/q/project:glusterfs+branch:master+topic:bug-1211863

Hopefully, it will be available in RHGS 3.2

Comment 12 Poornima G 2017-09-21 09:26:57 UTC
Lot os fixes have gone in 3.2 and 3.3 for the same:
1. md-cache improvements: 3.2
2. Negative lookup cache: 3.3
3. Parallel-readdir: 3.3

Hence will close this as fixed in 3.3.

Please test the performance on 3.3 and raise a different BZ for further perf improvements request.


Note You need to log in before you can comment on or make changes to this bug.