[Bug 1885730] Re: Bring back ondemand.service or switch kernel default governor for pstate - pstate now defaults to performance governor

[Colin Ian King]

@xnox,

one can detect the machine type from the DMI data (iff it is available
and reliable).

e.g. on my laptop:

sudo dmidecode  -s "chassis-type"
Notebook

on my desktop server:
sudo dmidecode  -s "chassis-type"

There are quite a few chassis-type, see https://www.dmtf.org/sites/default/files/standards/documents/DSP0134_2.7.1.pdf - page 36, section 7.4.1, table 16 "System Enclosure or Chassic Type".
*Note* some machines may have illegal data or omitted this information and it's DMI specific, so s390, arm64, ppc64el platforms most probably won't have this.


So some context. This performance governor was chosen for the boot default setting because it speeds up boot. I've performed some simple boot tests on an older generation Lenovo x220 (i3-2350M), timings are below for a 5.8 kernel:

Governor     Kernel   Userspace  EndBoot
PEFORFMANCE  2.15s    10.37s     12.52s
POWERSAVE    2.87s    18.49s     21.37s
ONDEMAND     2.41s    10.46s     12.87s

So PERFORMANCE saves a few tenths of a second over ONDEMAND, hence this
choice for boot. The setting thereafter is obviously more complex.  I've
compared the governor settings on the same laptop running idle, with 50%
CPU utilization and 100% CPU utilization.  I used powerstat to monitor
CPU power, CPU freq and C7 (deep sleep) residency.

1. Idle.
  - power utilization roughly the same in power usage, but powersave clocks the CPU lowest. Note that the CPU freq is cranked up when measurements are taken, so it's hard to get a correct freq. measurement.

2. 50% CPU busy.
   - performance consumes most power (as expected) and the CPU is running marginally faster than on-demand.

3. 100% CPU busy.
   - performance and on-demand are comparable in terms of power consumption and CPU frequency.

PERFORMANCE essentially runs the CPU at higher frequency all the time,
whereas ONDEMAND will scale up/down depending on the utilization.  I
believe the default should be ONDEMAND post boot as this is the most
flexible option and will provide power saving when the system is less
utilized.  If users want to burn power and they can opt-in to manually
setting to PERFORMANCE, but I think this should be opt-in rather than
the default setting for any class of machine.

Finally, if we don't want the userspace changes, we could default the
kernel to ONDEMAND and take a hit on slower boot performance.

To clarify I see the options as:

1. Boot with PERFORMANCE and fix the userspace to set ONDEMAND in the post-boot stage
2. Failing this, don't let userspace do anything smart post-boot and default to ONDEMAND

Attached are some data points I gathered with the 5.8 kernel.





** Attachment added: "LibreOffice |Calc spread sheet of boot timings and power measurements"
   https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1885730/+attachment/5402512/+files/2020-5.8-CPU-GOVERNOR-CONFIG.ods

-- 
You received this bug notification because you are a member of Ubuntu
Foundations Bugs, which is subscribed to systemd in Ubuntu.
https://bugs.launchpad.net/bugs/1885730

Title:
  Bring back ondemand.service or switch kernel default governor for
  pstate - pstate now defaults to performance governor

Status in linux package in Ubuntu:
  Confirmed
Status in systemd package in Ubuntu:
  Invalid
Status in linux source package in Focal:
  New
Status in systemd source package in Focal:
  Fix Released
Status in linux source package in Groovy:
  Confirmed
Status in systemd source package in Groovy:
  Invalid

Bug description:
  In a recent merge from Debian we lost ondemand.service, meaning all
  CPUs now run in Turbo all the time when idle, which is clearly
  suboptimal.

  The discussion in bug 1806012 seems misleading, focusing on p-state vs
  other drivers, when in fact, the script actually set the default
  governor for the pstate driver on platforms that use pstate.
  Everything below only looks at systems that use pstate.

  pstate has two governors: performance and powerstate. performance runs
  CPU at maximum frequency constantly, and powersave can be configured
  using various energy profiles energy profiles:

  - performance
  - balanced performance
  - balanced power
  - power

  It defaults to balanced performance, I think, but I'm not sure.

  Whether performance governor is faster than powersave governor is not
  even clear.
  https://www.phoronix.com/scan.php?page=article&item=linux50-pstate-
  cpufreq&num=5 benchmarked them, but did not benchmark the individual
  energy profiles.

  For a desktop/laptop, the expected behavior is the powersave governor
  with balanced_performance on AC and balanced_power on battery.

  I don't know about servers or VMs, but the benchmark series seems to
  indicate it does not really matter much performance wise.

  I think most other distributions configure their kernels to use the
  powersave governor by default, whereas we configure it to use the
  performance governor and then switch it later in the boot to get the
  maximum performance during bootup. It's not clear to me that's
  actually useful.

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1885730/+subscriptions