Google Proposes Multi-Generational LRU For Linux To Yield Much Better Performance

[nwildner]

I welcome their contributions, but it's mostly a function of self-interest. So, I don't let it color my overall views of those corporations.

You need to be a special kind of moron to think that contributions coming from companies are always on good heart... C'mon, it's 2021 and a lot of companies already contributed with Linux.

And double that with the fact that you are indeed being granted with better memory management for your self-interest: Use Linux on a daily basis.

[[email protected]]

Thank you. The latest patches are posted at the following link, in case anybody is interested in previewing or testing them.

https://linux-mm-review.googlesource...claim/+/1101/3

A new series is available: https://lore.kernel.org/linux-mm/[email protected]/

There is an improvement in buffered io performance under memory pressure!

[[email protected]]

This is awesome man, I greatly appreciate the attitude here!

Thank you. The latest patches are posted at the following link, in case anybody is interested in previewing or testing them.

https://linux-mm-review.googlesource.com/c/page-reclaim/+/1101/3

[geearf]

Hi, I'm one of the Google engineers who work on the multigenerational LRU. We'd be happy to work with you, if you are interested.

Please feel free to send any questions to [email protected], or comment on the latest doc: https://docs.google.com/document/d/1...ZlYoebTVM/edit

Quoting a Vulcan proverb: we are here to serve.

This is awesome man, I greatly appreciate the attitude here!

[cb88]

Hi, I'm one of the Google engineers <snip>

Quoting a Vulcan proverb: we are here to serve ADs.

There much more accurate!

[coder]

And Microsoft, and Oracle. While one might not agree with any particular corporate policy, the resources (i.e. money) they put into Linux should be welcomed.

I welcome their contributions, but it's mostly a function of self-interest. So, I don't let it color my overall views of those corporations.

Recall that Google was one of Firefox's biggest backers, before they had Chrome. Then, their support for it nearly evaporated. The same will happen with Linux, when Google eventually transitions Android onto Fuschia and their cloud onto ...whatever.

MS and Oracle are being equally pragmatic. MS realized that it was losing cloud market share, and Oracle was probably just looking for ways to save money by shedding its Solaris developers.

[[email protected]]

I'm running 5.12-rc2 with the LRU patches right now and nothing is obviously broken. What PTS tests should I run to best view performance improvements? I have an rc2 without the patches I want to compare side by side.

Thank you!!! Do feel free to ask on the kernel memory mailing list if you have any problem with the patchset. Or send emails to [email protected]. We'll do our best to help.

[[email protected]]

It would be nice if someone could post backports of this patch sets for previous kernels. So we can try them on sbc's which tend to be tight on ram.

Backports for 4.14, 4.19 and 5.4 will be available by the end of 2021H1. Would this timeline work for you? Thank you.

[[email protected]]

Very cool. Always room in my heart for better algorithms that are simpler to implement than inferior ones; and I will have to look at this paper for work now, since one of the problems we have is very similar to this problem.

Hi, I'm one of the Google engineers who work on the multigenerational LRU. We'd be happy to work with you, if you are interested.

Please feel free to send any questions to [email protected], or comment on the latest doc: https://docs.google.com/document/d/1...ZlYoebTVM/edit

Quoting a Vulcan proverb: we are here to serve.

[microcode]

It would be nice if someone could post backports of this patch sets for previous kernels. So we can try them on sbc's which tend to be tight on ram.

Well, it depends on your SBC, but maybe your kernel should be newer, instead of backporting performance improvements (though that's not to say you couldn't do it).