Linux 4.9 I/O Scheduler Benchmarks On A SSD: Noop vs. CFQ vs. Deadline

[phoronix]

Phoronix: Linux 4.9 I/O Scheduler Benchmarks On A SSD: Noop vs. CFQ vs. Deadline

Some Phoronix Premium readers had recently requested some fresh I/O scheduler benchmarks using the Linux 4.9 kernel, so here are those test results...

http://www.phoronix.com/scan.php?page=news_item&px=Linux-4.9-IO-Scheduler-SSD

[Guest]

Thanks for providing such benchmarks.

[wpupkin]

What about blk-mq ?

[peppercats]

What about blk-mq ?

this, I thought I/O schedulers were effectively dead now?

[FireBurn]

Why wasn't no-op in the last result? And which one was best overall?

[xevrem]

this, I thought I/O schedulers were effectively dead now?

Not quite. The push now from the I/O scheduler folks is how to keep throughput high while minimizing impact to responsiveness. BFQ typically wins this use-case, but wont beat CFQ for raw throughput, but then that isnt the point of BFQ.

[gamerk2]

Not quite. The push now from the I/O scheduler folks is how to keep throughput high while minimizing impact to responsiveness. BFQ typically wins this use-case, but wont beat CFQ for raw throughput, but then that isnt the point of BFQ.

Yep; with any scheduler, you are basically trying to optimize latency/throughput for your specific use case. Depending on workload, one scheduler could be far superior to another.

[samdraz]

Doesn't samsung 850 evo series have problem with queued trim(data loss)?

[FuturePilot]

Doesn't samsung 850 evo series have problem with queued trim(data loss)?

Yes, it's still blacklisted AFAIK.

[sligor]

Doesn't samsung 850 evo series have problem with queued trim(data loss)?

I don't see the "discard" ext4 mount option, so this will have no impact in these benchmarks.