Submit scheduler to Mainline Kernel?

[microchip8]

Hi,

Are there any plans to submit this I/O scheduler to the mainline kernel? I've had really good experience with it so far, both on SSDs and HDDs, and having it in the kernel will remove one more patch before compiling a new kernel for my system.

Thanks :)

[firelzrd]

Hi.

Thank you so much for your kind words about the I/O scheduler.
I really appreciate you taking the time to share your positive experience with it.
It’s great to hear that it’s working well for both your SSDs and HDDs.

Regarding submission to the mainline kernel, I haven’t seriously considered it until now, but I’ll definitely explore the possibility.
That said, I’m a bit nervous that kernel development has strict conventions, so I expect plenty of constructive criticism.
Still, if there’s interest from users like you, it might be worth the effort.

Would you mind sharing any specific use cases or benchmarks where this scheduler shines?
Feedback like that could help strengthen a potential proposal.

Thanks again for your support!

Best regards

[microchip8]

I haven't done any benchmarks. What I've noticed, though, is that my desktop is more responsive when I use it and at the same time remux a Blu-ray (60-70 GiB in size) to mkv format with MakeMKV running in the background. Previously, I was using Kyber on my SSDs but ADIOS makes it more responsive. Kyber on HDDs is a no-go, while ADIOS behaves the same on SSDs and HDDs. Especially when I'm writing the mkv to a HDD, non of the mainline I/O schedulers can guarantee a smooth response. I've noticed that if 'm watching something from the HDD using mpv while at the same time having MakeMKV write hammer the disk, the video does not "hang" for brief moments with ADIOS, while it does using BFQ or mq-deadline. That's my experience so far.

[tolga9009]

Hello,

I want to encourage you to upstream this scheduler to Linux. The review of 2000 lines of code will take some time and you can surely expect to go through multiple revisions. But please take it as an opportunity, both for learning about the process of upstream Linux development, as well as a career pushing option. If this makes it into Linux one day and any company finds out this the best scheduler for their workload, your career options can explode over night.

If there is harsh critic, which is absolutely possible - so what? The scheduler evidently works and even if it's not beautiful enough for upstream, there are still happy users. The usual process is like this: they review your code, tell you what's good and what isn't and they give you a chance to post another round of updated patches with the mentioned changes. Rinse and repeat until either your code gets accepted or rejected (e.g. due to unfixable issues).

Of course, to underline why ADIOS is important, you'd need some performance benchmarks. Your YouTube video itself is very impressive already. An additional table containing raw performance data can't hurt either.

Best of luck!

Cheers,
Tolga