tjh.dev
Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
docs: scheduler: Start documenting the EEVDF scheduler
Add some documentation regarding the newly introduced scheduler EEVDF.
Signed-off-by: Carlos Bilbao <carlos.bilbao.osdev@gmail.com>
Signed-off-by: Jonathan Corbet <corbet@lwn.net>
Link: https://lore.kernel.org/r/20240720002207.444286-2-carlos.bilbao.osdev@gmail.com
authored by
Carlos Bilbao
and committed by
Jonathan Corbet
602bce7e
e9c7acd7
+50
-4
+1
Documentation/scheduler/index.rst
+1
Documentation/scheduler/index.rst
+6
-4
Documentation/scheduler/sched-design-CFS.rst
+6
-4
Documentation/scheduler/sched-design-CFS.rst
···
8
8
1. OVERVIEW
9
9
============
10
10
11
-
CFS stands for "Completely Fair Scheduler," and is the new "desktop" process
12
-
scheduler implemented by Ingo Molnar and merged in Linux 2.6.23. It is the
13
-
replacement for the previous vanilla scheduler's SCHED_OTHER interactivity
14
-
code.
11
+
CFS stands for "Completely Fair Scheduler," and is the "desktop" process
12
+
scheduler implemented by Ingo Molnar and merged in Linux 2.6.23. When
13
+
originally merged, it was the replacement for the previous vanilla
14
+
scheduler's SCHED_OTHER interactivity code. Nowadays, CFS is making room
15
+
for EEVDF, for which documentation can be found in
16
+
Documentation/scheduler/sched-eevdf.rst.
15
17
16
18
80% of CFS's design can be summed up in a single sentence: CFS basically models
17
19
an "ideal, precise multi-tasking CPU" on real hardware.
+43
Documentation/scheduler/sched-eevdf.rst
+43
Documentation/scheduler/sched-eevdf.rst
···
1
+
===============
2
+
EEVDF Scheduler
3
+
===============
4
+
5
+
The "Earliest Eligible Virtual Deadline First" (EEVDF) was first introduced
6
+
in a scientific publication in 1995 [1]. The Linux kernel began
7
+
transitioning to EEVDF in version 6.6 (as a new option in 2024), moving
8
+
away from the earlier Completely Fair Scheduler (CFS) in favor of a version
9
+
of EEVDF proposed by Peter Zijlstra in 2023 [2-4]. More information
10
+
regarding CFS can be found in
11
+
Documentation/scheduler/sched-design-CFS.rst.
12
+
13
+
Similarly to CFS, EEVDF aims to distribute CPU time equally among all
14
+
runnable tasks with the same priority. To do so, it assigns a virtual run
15
+
time to each task, creating a "lag" value that can be used to determine
16
+
whether a task has received its fair share of CPU time. In this way, a task
17
+
with a positive lag is owed CPU time, while a negative lag means the task
18
+
has exceeded its portion. EEVDF picks tasks with lag greater or equal to
19
+
zero and calculates a virtual deadline (VD) for each, selecting the task
20
+
with the earliest VD to execute next. It's important to note that this
21
+
allows latency-sensitive tasks with shorter time slices to be prioritized,
22
+
which helps with their responsiveness.
23
+
24
+
There are ongoing discussions on how to manage lag, especially for sleeping
25
+
tasks; but at the time of writing EEVDF uses a "decaying" mechanism based
26
+
on virtual run time (VRT). This prevents tasks from exploiting the system
27
+
by sleeping briefly to reset their negative lag: when a task sleeps, it
28
+
remains on the run queue but marked for "deferred dequeue," allowing its
29
+
lag to decay over VRT. Hence, long-sleeping tasks eventually have their lag
30
+
reset. Finally, tasks can preempt others if their VD is earlier, and tasks
31
+
can request specific time slices using the new sched_setattr() system call,
32
+
which further facilitates the job of latency-sensitive applications.
33
+
34
+
REFERENCES
35
+
==========
36
+
37
+
[1] https://citeseerx.ist.psu.edu/document?repid=rep1&type=pdf&doi=805acf7726282721504c8f00575d91ebfd750564
38
+
39
+
[2] https://lore.kernel.org/lkml/a79014e6-ea83-b316-1e12-2ae056bda6fa@linux.vnet.ibm.com/
40
+
41
+
[3] https://lwn.net/Articles/969062/
42
+
43
+
[4] https://lwn.net/Articles/925371/