Guest post by Linux 2.6.37 has been released. This release includes several SMP scalability improvements for Ext4 and XFS, complete removal of the Big Kernel Lock, support for per-cgroup IO throttling, a network device based in the Ceph clustered filesystem, several Btrfs improvements, more efficient static probes, perf support to probe modules and listing of accesible local and global variables, image hibernation using LZO compression, PPP over IPv4 support, several networking microoptimizations and many other small changes, improvements and new drivers. You can read the full changelog as well.
just can’t wait for the Galbraith latency patch in 2.6.38
I like the most things they have done to kernel, particularly removal of Big Kernel Lock, or IO throttling. But I like some day to see a power manager as good as what OSX has.
OS X has piss poor power management in and of itself, it’s the Apple drivers and tweaks that allow it to be efficient. You can do this in Linux too, but it takes more work and some understanding of how ACPI works.
OS X does not have to support as much software. And when you buy a windows laptop the laptop manufacturer installs it’s “own” drivers to make it efficient.
So the Linux developers have a lot more work to do.
But if you are unhappy how much power your Linux-based system uses, I suggest running ‘powertop’. It will help you with all of that.
Edited 2011-01-06 17:02 UTC
Based on what evidence is it ‘piss poor’ or is this more ‘speaking out of your ass’ rather than using empirical data?
The efficiency in terms of power management has as much to do with the kernel as it has with the user land applications and how they’re written. If an application is polling non-stop for a device to be inserted then it won’t matter how efficient the kernel is given that an application in user space is constantly waking the CPU from its low power state.
Well, this one will be most talked about I guess: http://www.kroah.com/log/linux/android-kernel-problems.html
Uh, why would it be talked about? That’s not new to this release – it happened in 2.6.33, over a year ago now, and all conversation occurred back then…
Edited 2011-01-05 20:31 UTC
I thought they had removed this a while ago. What actually happened is that they gradually lessened the impact of The Big Kernel lock, allowing more and more of the kernel to be run in parallel. The only think it controls in modern 2.6 kernels are a hand full of drivers and obscure file systems. They fully removed the Big Kernel lock and replaced it with mutexes. So these parts that were not able to be run in parallel, are still not able to run in parallel, but one obscure driver shouldn’t block a different obscure driver anymore. For what that’s worth.
Still a pretty cool benchmark for the kernel. Oh, I remember the taunts of those who criticised the linux 2.0 kernel for its SMP. Nice to lay those to rest.
That combined with the recent scheduler changes in 2.6.38 and lots of changes to the kernel-threads, some tickless changes and ext4 scalability improvements and removal of barriers in the block subsystem should hopefully improve performance for a bit.
Edited 2011-01-06 17:17 UTC
I’ll stick with 2.4, much faster thanks.
I wonder what you use it for.
wrong reply
Edited 2011-01-06 17:00 UTC