PDS-MQ CPU Scheduler Revised For The Linux 4.17 Kernel With Minor Optimizations

[phoronix]

Phoronix: PDS-MQ CPU Scheduler Revised For The Linux 4.17 Kernel With Minor Optimizations

Alfred Chen announced this week the release of PDS-mq 0.98s, his latest patch-set of this CPU scheduler against the Linux 4.17 upstream code-base and includes minor optimization work and bug fixes...

http://www.phoronix.com/scan.php?page=news_item&px=PDS-MQ-Linux-4.17

[busukxuan]

Typo in 2nd paragraph

Phoronix: PDS-MQ CPU Scheduler Revised For The Linux 4.17 Kernel With Minor Optimizations

The PDS scheduler stands for the "Priority and Deadline based Skiplist multiple queue scheduler" that is derived from Con Kolivas' former BFS scheduler with Variable Run Queue (VRQ) support. PDS design principles are to be a cimple CPU process scheduler yet efficient and scalable. PDS-mq differs from Con Kolivas' current MuQSS scheduler.


http://www.phoronix.com/scan.php?pag...-MQ-Linux-4.17

[tildearrow]

Typo:

are to be a
cimple CPU process

[hax0r]

There is a port of FreeBSD's ULE scheduler to Linux available at https://github.com/JBouron/linux, I recently read about it in this paper https://www.usenix.org/system/files/...c18-bouron.pdf. Would be interesting to have it benchmarked.

ULE may cause starvation, even when executing a single application with identical threads, but this starvation may actually lead to better application performance for some workloads. The more complex load balancing mechanism of CFS reacts more quickly to workload changes, but ULE achieves better load balance in the long run.

[halo9en]

I agree a benchmark of both PDS and ULE compared to mainline schedulers would be interesting.

[jwh7]

Fwiw, I've been running PDS w/ BFQ (mainline plus patches) for a long time without issue; ran BFS then MuQSS for a while before that. This is with Arch on an old x64 Athlon X2 desktop (LXQT usually) and older x86 UP netbook (LXDE); definitely works well.