Linux to get a 10x speed boost?

Ashton Mills
18 March 2011, 3:13 PM


In Linux land a single patch to the kernel isn't usually ground-breaking, but a small patch by kernel developer Mike Galbraith has even Linus Torvalds excited.


In the past, Linux has been lambasted for being too server-centric, with many new features and innovations focused on the enterprise, rather than on the desktop. One of the common complaints here is how the default Linux kernel is tuned for throughput (something servers love) rather than latency (something desktop users love) - and they are two sides of the same coin, with one coming at the expense of the other. Latency is important for desktops because it provides a more responsive desktop environment for the user - moving windows around, scrolling a web page for example - especially while under load.


Mike Gailbraith's new Linux patch promises to cut desktop latency by a factor of ten.

Which is why the small patch contributed by Mike Galbraith is so important, as it promises to cut desktop latency by a factor of ten. The official explanation of how the patch works is loaded with developer jargon, but the skinny seems to be that it takes advantage of CPU task groups and group scheduling to improve desktop interactivity on the fly while under load. Granted, your system isn't often likely to be under heavy load, unless you do video encoding or software development, but according to Linus Torvalds, he even found web pages loading faster, so it may have benefits for general system usage.

Scheduling is a vast and complex science that poses issues for Linux and Windows alike, so for such a small patch to make such a big difference it's no wonder Linus is happy with the result. The patch is included in the latest 2.6.38 kernel, which likely won't make it in time to be to be included in the next release of Ubuntu (Ubuntu 11.04, codenamed Natty Narwhal), due at the end of April. However, Ubuntu developers often patch the Linux kernel with extra driver support and features, so we may see this patch backported to the 2.6.37 kernel Natty is based on.


Post your comment



Comments

RSS feed Email alert

Me In Oz (User):

It took 20 years for the penny to drop?
And 10 x faster? ...... Lets see someone post some facts and not just claims!

".... he even found web pages loading faster, so it may have benefits for general system usage..."
That's hilarious :)

18 March 2011, 3:47 PM (2 years ago)report abuse Send to a friend reply

petert (Advanced Forumologist):

Facts are like statistics . . . they are in the eye of the beholder. Take as an example the "getthefacts" site of MS, which could hardly be described as at factual at all!

18 March 2011, 4:27 PM (2 years ago)report abuse Send to a friend reply

Tin (User):

Quoting Me In Oz:
It took 20 years for the penny to drop?


Nope. Linux has had a choice of schedulers for as long as I've been fiddling - which is now about 12 years.

Quoting Me In Oz:
And 10 x faster? ...... Lets see someone post some facts and not just claims!

If this is the patch I think it is (which is months old, BTW), then the "ten times faster" is about the wait times on IO, not about processing power or how fast things actually load.
The end result is the time between clicking an icon and the kernel actually starting to act on the click is reduced. Combine this by the hundreds of times a minute you're waiting for the kernel to act on things, and it adds up to a noticeable difference.

18 March 2011, 5:43 PM (2 years ago)report abuse Send to a friend reply

ss-rotel (User):

this is all kinda relative, as usually ppl that use this, (unless you're die hard, and probably wont read this site anyway), will have it running on really old machine, that would strugle to run XP, so they are too tight to buy XP, and as such, to upgrade to a faster machine anyway ;)



18 March 2011, 5:06 PM (2 years ago)report abuse Send to a friend reply

petert (Advanced Forumologist):

And you can cite research that proves any one of your contentions?

19 March 2011, 12:54 PM (2 years ago)report abuse Send to a friend reply

Me In Oz (User):

Quoting petert:
And you can cite research that proves any one of your contentions?

I'll just let you answer this one petert:"

" ... Facts are like statistics . . . they are in the eye of the beholder." - petert




19 March 2011, 1:05 PM (2 years ago)report abuse Send to a friend reply

petert (Advanced Forumologist):

Touche! Well played ;-)

19 March 2011, 1:19 PM (2 years ago)report abuse Send to a friend reply

anonymous user Anonymous user