You are not logged in.
- Topics: Active | Unanswered
#1 2016-07-03 17:49:42
- graysky
- Wiki Maintainer
- From: :wq
- Registered: 2008-12-01
- Posts: 10,597
- Website
Booting with scsi_mod.use_blk_mq=1 ... poor IO latency [solved]
When using scsi_mod.use_blk_mq=1, have you noticed poor latency when disk IO is in demand? On a SATA3 HDD (Sragate Barracuda ST3000DM001), I am currently restoring a 500G backup via rsync (remote server on the LAN) and I can observe a massive performance hit as I use IO to the same drive for small copy tasks. I did not experience this under the cfq IO scheduler. I also noticed it when zeroing out some free space on the same HDD using the dm-crypt wipte method. Others care to comment?
EDIT: seems this is to be expected.
Last edited by graysky (2016-08-07 13:33:00)
CPU-optimized Linux-ck packages @ Repo-ck • AUR packages • Zsh and other configs
Offline
#2 2016-08-02 11:18:50
- nefelim4ag
- Member
- From: Minsk
- Registered: 2014-01-05
- Posts: 3
- Website
Re: Booting with scsi_mod.use_blk_mq=1 ... poor IO latency [solved]
@graysky, blk-mq disable io schedulers, so it's expected on HDD
Offline
#3 2016-08-07 11:04:34
- graysky
- Wiki Maintainer
- From: :wq
- Registered: 2008-12-01
- Posts: 10,597
- Website
Re: Booting with scsi_mod.use_blk_mq=1 ... poor IO latency [solved]
Ah! Seems like another method to enabled it for only SSDs would be needed then... I struck out googling for this...
EDIT: https://mahmoudhatem.wordpress.com/2016 … el-blk-mq/
Last edited by graysky (2016-08-10 05:36:23)
CPU-optimized Linux-ck packages @ Repo-ck • AUR packages • Zsh and other configs
Offline