Improving the performance of the BFQ I/O scheduler

Posted Apr 1, 2019 15:31 UTC (Mon) by wazoox (subscriber, #69624)
Parent article: Improving the performance of the BFQ I/O scheduler

Would also be nice to include the noop scheduler in the comparisons, for reference. Well not only for reference, I actually always use noop on servers :)

Improving the performance of the BFQ I/O scheduler

Posted Apr 1, 2019 16:09 UTC (Mon) by axboe (subscriber, #904) [Link] (2 responses)

noop no longer exists, that went away with the legacy IO stack. The equivalent now is 'none'.

That aside, I do agree, it would have been prudent to have all graphs include none/mq-deadline/kyber instead of picking and choosing a particular one for each test. Selectively including a particular scheduler for each comparison can seem a bit shady.

Improving the performance of the BFQ I/O scheduler

Posted Apr 1, 2019 20:16 UTC (Mon) by paolo (guest, #126832) [Link] (1 responses)

For brevity, I showed results only for the best performing other I/O schedulers. Full results for these and other tests can be found here:
https://algo.ing.unimo.it/people/paolo/disk_sched/results...

Improving the performance of the BFQ I/O scheduler

Posted Apr 1, 2019 20:24 UTC (Mon) by axboe (subscriber, #904) [Link]

That's very useful, thanks Paolo!