Five Years Of Linux Kernel Benchmarks: 2.6.12 Through 2.6.37

[birdie]

cb88,

Why do you need to run the latest and greatest software on your ancient hardware? Run kernel 2.4 which works very well. And I doubt you'll need to run any modern desktop software like word processors or web browsers. You see, the web has evolved a lot for the last ten years. Old web browsers won't be able to render correctly most modern popular web sites.

So, either run old software on your old hardware or find another thing to do.

After all people are not complaining they cannot ride horses in modern cities. The world is evolving rapidly. Live with that or perish.

[Petteri]

Heck, you can run recent linux kernel and full-featured userspace with all the eye-candy in mobile phones. Or in basic wireless routers with 8MB ram and 2 MB flash (without the eye-candy obviously).
Linux kernel getting more bloatet isn't huge problem (or even problem at all).

[aussiebear]

cb88,

Why do you need to run the latest and greatest software on your ancient hardware? Run kernel 2.4 which works very well. And I doubt you'll need to run any modern desktop software like word processors or web browsers. You see, the web has evolved a lot for the last ten years. Old web browsers won't be able to render correctly most modern popular web sites.

So, either run old software on your old hardware or find another thing to do.

After all people are not complaining they cannot ride horses in modern cities. The world is evolving rapidly. Live with that or perish.

I think you missed the point...They were referring to efficiency. Not age.

There seems to be a weird habit of programmers bloating things up for no reason other than to "take advantage of new hardware". ie: no longer care about efficiency, and features become the priority...It results in sloppy programming practices.

To be honest, if you have a choice between bloat or "lean and mean"; which would you go for? Most people would go for the latter when there is an option to do so.

Then again, not all things modern have been beneficial to humanity...It has even caused consequences which many rather not admit. So we end up treating symptoms.

[smitty3268]

There seems to be a weird habit of programmers bloating things up for no reason other than to "take advantage of new hardware". ie: no longer care about efficiency, and features become the priority...It results in sloppy programming practices.

I really don't think you'll find very much of that among kernel developers. That's why they are kernel developers and not writing websites or desktop apps, and if you read through the kernel mailing lists i think you'll find that there isn't a whole lot of tolerance there for people who write bad code.

[L33F3R]

god help the linux kernel

[MaestroMaus]

I really don't think you'll find very much of that among kernel developers. That's why they are kernel developers and not writing websites or desktop apps, and if you read through the kernel mailing lists i think you'll find that there isn't a whole lot of tolerance there for people who write bad code.

+1

The kernel is doing well so far. Most of the bad code and bloat-ware is found in other parts of the software stack.

[Jimbo]

I really don't think you'll find very much of that among kernel developers. That's why they are kernel developers and not writing websites or desktop apps, and if you read through the kernel mailing lists i think you'll find that there isn't a whole lot of tolerance there for people who write bad code.

+1

The problem is not supporting old hardware or bad kernels developers. As linux says, the problem is that a lot new functionality has been added recently without time to getting stabilized, and this could be dangerous on solid environments: KMS (lots of code), apparmor, responsability patches, VFS, schedulers...

I believe performance is not currently being compromised, if u try ext4 or xfs with nobarrier mount option, you should find that recent kernels beats ext3 benchmarks easily.

[TemplarGR]

I believe Michael's methodology was flawed. I am surprised none mentioned that these tests were run inside a virtual machine. There are no useful conclusions from these tests:

1) Host was using almost the latest kernel, so the virtual machine used all current improvements anyway.

2) Userland from what i understand was still from 2005. I believe in many cases a more up-to-date userland could levereage more from recent kernels in some of the tests.

A quick way to test this, is using a 2005 era machine, and put Fedora Core 4, benchmark it, and then install Fedora 14(or simply compile the latest kernel if you wish only kernel improvements) and benchmark it again. I strongly believe there will be many differences in results...

[Guest]

I believe Michael's methodology was flawed. I am surprised none mentioned that these tests were run inside a virtual machine.

That's maybe because Phoronix methodoloy is always flawed and nobody takes these results serious?

At least not kernel devs who knows something about testing methodologies

[mtippett]

I believe Michael's methodology was flawed. I am surprised none mentioned that these tests were run inside a virtual machine. There are no useful conclusions from these tests:

1) Host was using almost the latest kernel, so the virtual machine used all current improvements anyway.

2) Userland from what i understand was still from 2005. I believe in many cases a more up-to-date userland could levereage more from recent kernels in some of the tests.

Almost all old kernels would exist in virtual machines in corporate environments these days. The expensive Linux application that you bought or got built 10 years ago that is only supporting and working on Red Hat 7.1 won't be running on the same hardware. That original hardware would have expired a few years after deployment. Most corporate environments then will move those installations to a VM.

Particularly for the older kernels, it's closer to a real production scenario then most would believe.