0

I am trying to measure the impact of Deadline IO scheduler on the performance of Block IO.

My setup includes:

  • Kernel Version: 4.4.154
  • Storage device: eMMC 64GB
  • Distribution: VERSION="18.04.4 LTS (Bionic Beaver)

I have tried the The Andrew Morton Interactive Workload (AMIW) as suggested in https://lwn.net/Articles/10874/ i.e. Flush lots of stuff to disk (I start a dbench xxx, or do a dd if=/dev/zero of=test_file bs=64k), and then time a cat dir/*.c where dir/ holds lots of source files.

But I see more or less the same amount of time taken for read for [cfq], [deadline] and [noop].

Why is that?

What test case should I use to see a substantial improvement in performance?

nishad kamdar
  • 73
  • 6
  • I suggest you have a look there https://www.phoronix.com/scan.php?page=article&item=linux5-io-sched&num=1 in order to see the different tests achieved. Then you can even download the full testsuite if interested https://www.phoronix-test-suite.com/ in order to run it yourself on your own hardware. – MC68020 Sep 16 '20 at 10:24
  • 1
    And BTW... there is no much thing to optimize regarding the regular flow of a dd command. – MC68020 Sep 16 '20 at 10:29
  • 1
    I found this script measuring performance of IO schedulers. https://github.com/m1d1/sched-bench/blob/master/sched-bench.sh – Michael D. Sep 16 '20 at 10:31

0 Answers0