This bug is missing log files that will aid in diagnosing the problem.
While running an Ubuntu kernel (not a mainline or third-party kernel)
please enter the following command in a terminal window:
apport-collect 1812569
and then change the status of the bug to 'Confirmed'.
If, due to the nature of the issue you have encountered, you are unable
to run this command, please add a comment stating that fact and change
the bug status to 'Confirmed'.
This change has been made by an automated script, maintained by the
Ubuntu Kernel Team.
** Changed in: linux (Ubuntu)
Status: New => Incomplete
--
You received this bug notification because you are a member of Kernel
Packages, which is subscribed to linux in Ubuntu.
https://bugs.launchpad.net/bugs/1812569
Title:
Default kernel "cfq" I/O scheduler results in worse rotating disk
performance than necessary
Status in linux package in Ubuntu:
Incomplete
Bug description:
1) The release of Ubuntu you are using, via 'lsb_release -rd' or System ->
About Ubuntu
Ubuntu 18.04.1 LTS
2) The version of the package you are using, via 'apt-cache policy pkgname'
or by checking in Software Center
linux-image-generic 4.15.0.43.45
3) What you expected to happen
HDD performance is not unnecessary slow by default because the wrong "cfq" IO
scheduler is used
4) What happened instead
HDD performance is slow compared to other systems where other IO scheduler
like "deadline" or "noop" is used. Even random OOM killings occur because swap
performance is too low.
Hi all,
About a year ago I moved from CentOS 7 to Ubuntu 18.04 on a couple of
older desktop and laptop machines. One of the things I noticed after
the switch, was that disk IO performances seemed to be degraded, and
the machines often became unresponsive with lot of IO load, like
copying files or swapping when a lot of memory was used. People on
other Ubuntu machines with HDD that I know or use the mentioned
machines, also complained about annoying system lockups and hangs.
Degrading IO performance could be caused by many things, like drivers,
wrong block size aligning, heavier desktop environment and
applications and browsers and web applications that become heavier in
RAM and IO usage. I tried several things like using lighter XFCE
environment, switching from Firefox to Chrome, adding adblockers and
webtracker blockers, and using a tab suspension plugin, to lower
memory usage. But low performance issues still remained and it
appeared that especially swapping became terribly slow and even the
kernel OOM killer started killing random applications, something I
haven't experienced before on desktop machines.
After a while I found out that on the old CentOS 7 installs, the disk
IO scheduler was set to "deadline", while on the Ubuntu 18.04
installs, it is "cfq". So a month ago on three Ubuntu 18.04 I switched
the scheduler from "cfq" to "noop". Guess what, after a month I can
conclude that the machines turned to be much more responsive, no
strange OOM killings anymore, and even when significant swap is used,
the machine is still responsive and workable. I set it on other
peoples computers as well and also there, complaints about slowness
disappeared.
Could it be that
1. the CFQ schedulers, that stands for "Complete Fairness Queueing", is
causing all processes to use disk IO randomly simultaneously, and that is
causing heavy seeking activity on old-fashion rotating HDDs, causing it to
severely lower the IO throughput?
2. it is better to by default use the "deadline" or "noop" IO scheduler for
HDDs, by putting the following in /etc/udev/rules.d/50-scheduler.rules:
ACTION=="add|change", KERNEL=="sd[a-z]",
ATTR{queue/rotational}=="1",ATTR{queue/scheduler}="noop"
to gain much better IO performance in general for "older" systems using
HDDs.
3. is this something others experience as well, although I think there is no
'placebo' effect on my side. Is it wise to test with some benchmark?
Systems involved that I tested:
1. AMD Athlon(tm) II X4 640 Processor, 8GB RAM, Western Digital Blue 7200rpm
1TB disk (WDC WD10EZEX-00BN5A0)
2. Acer Aspire R3-471T-33NP laptop, Intel(R) Core(TM) i3-4005U CPU, 4GB RAM,
Seagate Laptop Thin HDD 5400rpm 500GB disk (ST500LT012-1DG142)
3. Acer Veriton L4630G desktop, Intel(R) Core(TM) i5-4440S CPU, 8GB RAM,
Western Digital Red 5400rpm 3TB disk (WDC WD30EFRX-68EUZN0)
These machines are all installed with LVM and amount of swap that
equals the amount of RAM, contrary the Ubuntu defaults of configuring
plain disk partitions and only 950MB of swap. This mainly because I
want the flexibility of LVM when I need to change disk configurations
later on, and I want to be able to use Hibernation (suspend to disk)
Typical usage: browsing websites, emailing using Thunderbird, using
LibreOffice documents, occasional copying over large amounts of files
to/from external USB disks, occasional image and video editing.
I see my more modern Thinkpad T450s laptop with SSD with Ubuntu 18.04
is also using CFQ, but there I never experience IO slowness, even when
GBs of swap are in use. Probably CFQ works fine with SSDs that have
much lower seek times.
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1812569/+subscriptions
--
Mailing list: https://launchpad.net/~kernel-packages
Post to : kernel-packages@lists.launchpad.net
Unsubscribe : https://launchpad.net/~kernel-packages
More help : https://help.launchpad.net/ListHelp
