[linux-azure] Commit To Improve NVMe Performance
Bug #1819689 reported by
Joseph Salisbury
This bug affects 1 person
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
linux-azure (Ubuntu) |
Fix Released
|
Undecided
|
Unassigned | ||
Xenial |
Fix Released
|
Undecided
|
Unassigned | ||
Cosmic |
Fix Released
|
Undecided
|
Unassigned |
Bug Description
In bug 1818138, the config option CONFIG_NO_HZ_FULL was requested to be enabled in the linux-azure kernels(currently 4.15 and 4.18) in order to increase NVME disks performance.
To achieve the peak IOPs we are seeing in testing, the following commit is also needed:
7ac257b862f2c (“blk-mq: remove the request_list usage”)
This commit cleans up unused code in block-mq (mq uses pre-allocated tags to allocate request, not request_list as it's used only in legacy queue). The commit saves a bunch of rcu locks on I/O queuing path, which increases the achievable IOPs seen in testing.
tags: | added: patch |
Changed in linux-azure (Ubuntu Cosmic): | |
status: | New → Fix Committed |
Changed in linux-azure (Ubuntu Xenial): | |
status: | New → Fix Committed |
To post a comment you must log in.
Hi, Joe.
Can you confirm the patch is also necessary for 4.15?
I backported the commit to xenial/linux-azure and it just required a bit of fixing due to context differences. I also built a test kernel if you want to test it:
https:/ /kernel. ubuntu. com/~mhcerri/ azure/xenial- linux-azure- 4.15.0- 1042.46+ lp1819689/