[solved] Changing i/o schedulers, good or bad idea?

midian · 2023-11-12 14:30:55

Hello,

for a few years I've been using the udev rules at https://forums.unraid.net/topic/98926-c … our-disks/

There were benchmarks for the different use cases with kernel 5.x and it seemed like a good idea back then.

Now I was wondering if this could still be useful today, or if the default i/o scheduler is good enough to manage the different storage media?

Last edited by midian (2023-11-15 18:10:08)

ewaller · 2023-11-12 15:49:51

I would not bother unless I had a very specific use case.

midian · 2023-11-13 07:05:41

Ok, thanks

c00ter · 2023-11-13 14:39:44

@ewaller, I'm denser than the average bear and also use the same udev i/o rules. Your reply to @midian indicates to me that it is now an unnecessary practice unless for "a very specific use case."

So that means, no, udev i/o rules are NOT generally needed any more (exception noted). Is this statement correct?

I wish I wasn't so on/off, yes/no, chocolate//vanilla, i/o, personality flawed, but @ age 71 I don't suppose that's ever going to change. Except pending one future event, anyway. Yes/no replies are greatly appreciated. Thank you.

Last edited by c00ter (2023-11-13 14:40:38)

ewaller · 2023-11-13 16:42:31

c00ter wrote:

@ewaller, I'm denser than the average bear and also use the same udev i/o rules. Your reply to @midian indicates to me that it is now an unnecessary practice unless for "a very specific use case." ... Yes/no replies are greatly appreciated. Thank you.

Not all answers are binary. General purpose desktop systems have different needs than a real time safety critical system which has different needs than a high performance database server. Critical real-time systems cannot be late. ever. Guaranteed timing might mandate accepting a lower overall throughput. A server probably wants the best overall throughput, but it is not a mission failure if -- occasionally -- it takes longer to do something than expected.

For a desktop system with an SSD, I think one would have a hard time perceiving the difference. I could be wrong. I still don't think it worth the effort.
Did you have a reason for doing this in the first place? Can you determine that the performance is different?

c00ter · 2023-11-14 15:30:22

The only reason I have it now is because I started using it several years ago in a different computer and the exact reason why is now lost in the fog. Perhaps from this wiki section on Improving Performance.: https://wiki.archlinux.org/title/Improv … _scheduler and used the 60-ioschedulers.rules example since it made sense, at least at that time. I had/have HDD & SATA SSDs per machine, and the example addressed them both.

No, I have not run any tests. I know we have the Phoronics test suite available but, truthfully, this 9th Gen. machine runs plenty fast enough for my daily desktop use. I'll test run with/without the rules under various loads, but I would bet you are correct about the perceived difference. I have used Linux for a time, but only at the desktop machine level. I *do* thank you very much for your easily understandable response explaining the different use cases to consider. It's a good reminder of why I love Linux and why I love Arch.

Last edited by c00ter (2023-11-14 15:33:37)

Arch Linux

#1 2023-11-12 14:30:55

[solved] Changing i/o schedulers, good or bad idea?

#2 2023-11-12 15:49:51

Re: [solved] Changing i/o schedulers, good or bad idea?

#3 2023-11-13 07:05:41

Re: [solved] Changing i/o schedulers, good or bad idea?

#4 2023-11-13 14:39:44

Re: [solved] Changing i/o schedulers, good or bad idea?

#5 2023-11-13 16:42:31

Re: [solved] Changing i/o schedulers, good or bad idea?

#6 2023-11-14 15:30:22

Re: [solved] Changing i/o schedulers, good or bad idea?

Board footer