Kernel autogroup patch

[naviathan]

Any projections on when the new kernel patch for group scheduling will be available for Mint? I realize it's only just been released into the git tree, but I figured it couldn't hurt to ask.

[JasonLG]

Any projections on when the new kernel patch for group scheduling will be available for Mint? I realize it's only just been released into the git tree, but I figured it couldn't hurt to ask.

Mint uses the Ubuntu version. So when they get it we get it. Rumor has it maybe as soon as Ubuntu 11.04.

[naviathan]

Is there a beta kernel in the repos that might get us the new patch sooner than that? 11.04 isn't far down the road fortunately. This patch claims massive performance improvements, which kills me given that Linux already runs circles around Windows. What's M$'s next excuse going to be?

[JasonLG]

Is there a beta kernel in the repos that might get us the new patch sooner than that? 11.04 isn't far down the road fortunately. This patch claims massive performance improvements, which kills me given that Linux already runs circles around Windows. What's M$'s next excuse going to be?

It's not even in the mainline kernel yet. But anythings possible, you might be able to find a compiled kernel some where that includes the patch and is compatible with Ubuntu/Mint if you look around?

As for M$ is concerned IMHO they really haven't made any real strides technically since the NT platform was released. Windows 7 really isn't much more than a kludged together patched XP with a shinny new wrapper. They prefer to compete by imitating and litigating rather than innovating these days.

[Susu]

Have a look: http://www.webupd8.org/2010/11/alternat ... patch.html

[naviathan]

Is there a beta kernel in the repos that might get us the new patch sooner than that? 11.04 isn't far down the road fortunately. This patch claims massive performance improvements, which kills me given that Linux already runs circles around Windows. What's M$'s next excuse going to be?

It's not even in the mainline kernel yet. But anythings possible, you might be able to find a compiled kernel some where that includes the patch and is compatible with Ubuntu/Mint if you look around?

As for M$ is concerned IMHO they really haven't made any real strides technically since the NT platform was released. Windows 7 really isn't much more than a kludged together patched XP with a shinny new wrapper. They prefer to compete by imitating and litigating rather than innovating these days.

Agreed, although I have to say, Win7 does run better on my netbook than XP ever did. Of course Mint blows them both away, thank you for dual booting! Maybe Win8 will finally be a worthy adversary given that M$ seems to finally be taking Linux seriously. It's no longer a techy or sysadmin OS.

[naviathan]

Have a look: http://www.webupd8.org/2010/11/alternat ... patch.html

Nice! I have to give this a go on my netbook.

Edit: I find it interesting that although the average latency is lower with Lennart's fix, the Max Latency is considerably higher than both the vanilla kernel with and without the patch.

[JasonLG]

Have a look: http://www.webupd8.org/2010/11/alternat ... patch.html

Nice! I have to give this a go on my netbook.

Edit: I find it interesting that although the average latency is lower with Lennart's fix, the Max Latency is considerably higher than both the vanilla kernel with and without the patch.

I've applied the non-patch hack to both a netbook running 32bit Mint 9 and a laptop running 64bit Mint 9. I don't notice much a difference on the netbook but on the laptop it seemed a little snappier when I was browsing the web while ripping a DVD with Handbrake. Totally non-scientific observation on my part, it could be a placebo effect, we'll see.

[naviathan]

I've applied the non-patch hack to both a netbook running 32bit Mint 9 and a laptop running 64bit Mint 9. I don't notice much a difference on the netbook but on the laptop it seemed a little snappier when I was browsing the web while ripping a DVD with Handbrake. Totally non-scientific observation on my part, it could be a placebo effect, we'll see.

There's an awful lot of videos and numbers out there from respected developers and users saying that this works beautifully and makes a dramatic difference. Now, that's not to say that you're not seeing a difference, but maybe the implementation isn't working on your netbook because something is missing or not configured appropriately?

[JasonLG]

I've applied the non-patch hack to both a netbook running 32bit Mint 9 and a laptop running 64bit Mint 9. I don't notice much a difference on the netbook but on the laptop it seemed a little snappier when I was browsing the web while ripping a DVD with Handbrake. Totally non-scientific observation on my part, it could be a placebo effect, we'll see.

There's an awful lot of videos and numbers out there from respected developers and users saying that this works beautifully and makes a dramatic difference. Now, that's not to say that you're not seeing a difference, but maybe the implementation isn't working on your netbook because something is missing or not configured appropriately?

I suspect that since I don't do any CPU intense multitasking on the netbook that is why I didn't notice much of a difference. That being said with the laptop Handbrake is generally a CPU hog using 100% of both cores while ripping a DVD. Usually it makes for a little sluggish web browsing but with the hack I ran Handbrake, VLC and Chromium with minimal slow down in chromium and little to no frame drops in VLC.

[naviathan]

I suspect that since I don't do any CPU intense multitasking on the netbook that is why I didn't notice much of a difference. That being said with the laptop Handbrake is generally a CPU hog using 100% of both cores while ripping a DVD. Usually it makes for a little sluggish web browsing but with the hack I ran Handbrake, VLC and Chromium with minimal slow down in chromium and little to no frame drops in VLC.

That's what I'm talking about. I can't wait to see how my quad core desktop handles with this. Maybe the netbooks don't show the performance increase as much because they're neither hyperthreaded or dual core. Could just be an effect of a single core no frills processor.

[JasonLG]

I suspect that since I don't do any CPU intense multitasking on the netbook that is why I didn't notice much of a difference. That being said with the laptop Handbrake is generally a CPU hog using 100% of both cores while ripping a DVD. Usually it makes for a little sluggish web browsing but with the hack I ran Handbrake, VLC and Chromium with minimal slow down in chromium and little to no frame drops in VLC.

That's what I'm talking about. I can't wait to see how my quad core desktop handles with this. Maybe the netbooks don't show the performance increase as much because they're neither hyperthreaded or dual core. Could just be an effect of a single core no frills processor.

My netbook has a 1.66GHz Atom N450 in it, it has one physical core but is hyperthreaded so it shows up as a dual core.

[naviathan]

Ok so you have a better netbook than I do. I'm still running the old N280 1.66GHz.

[Susu]

N270/280 have hyperthreading, too...

[naviathan]

N270/280 have hyperthreading, too...

So they do. I always thought they were single core single thread. Now I'm a bit disappointed in the performance...lol

[JasonLG]

N270/280 have hyperthreading, too...

So they do. I always thought they were single core single thread. Now I'm a bit disappointed in the performance...lol

From what I understand there isn't much difference between the N2xx and N4xx lines. I'n pretty happy with my N450, I did do a little tweaking to get it as fast as I could though. Have you looked into tweaking the file system and such to squeeze every little bit out of it.

[naviathan]

What kind of tweaks? I've mostly just done a load and go as far as linux is concerned. Win7 on the other hand I had to tweak a bit by shutting down unnecessary services and such.

[piratesmack]

Con Kolivas (BFS developer) has blogged about this patch
http://ck-hack.blogspot.com/2010/11/cre ... mment.html

Interesting read

[naviathan]

That is an interesting read. I wonder why he sees regressions where everyone else is seeing smoother overall operations?

[JasonLG]

What kind of tweaks? I've mostly just done a load and go as far as linux is concerned. Win7 on the other hand I had to tweak a bit by shutting down unnecessary services and such.

There are a few things you can do.
Here's what I've done.

Using bum(BootUp-Manager) you can disable unused services.
you can install bum with this command.

Code: Select all

sudo apt-get update && sudo apt-get install bum

I have disabled everything except.
avahi-daemon
pulseaudio
saned (if you don't have a scanner you can disable this)
cups (if you don't have a printer you can disable this)

With the startup applications editor I've disabled everything except these.
Certificate and Key Storage
Network Manager
Power Manager
Print Queue Applet (again if you don't have a printer you can disable this)
PulseAudio Sound System
Secret Storage Service
User folders update

I also change the swappiness to 10 by editing sysctl.conf
To do that enter this into a terminal.

Code: Select all

gksudo gedit /etc/sysctl.conf

Put this at the end of the file and save and close it.

Code: Select all

vm.swappiness=10

I then tweak the ext4 file system a little.
I first enter this in a terminal. replacing "*" with the number of the drive that you want to tune(for example sda1)

Code: Select all

sudo tune2fs -o journal_data_writeback /dev/sda*

I then add this to my fstab by entering this into a terminal.

Code: Select all

gksu gedit  /etc/fstab

and add data=writeback,noatime,nodiratime in the proper place(s)
It should look something like this. Here is what my fstab looks like.

Code: Select all

# /etc/fstab: static file system information.
#
# Use 'blkid -o value -s UUID' to print the universally unique identifier
# for a device; this may be used with UUID= as a more robust way to name
# devices that works even if disks are added and removed. See fstab(5).
#
# <file system> <mount point>   <type>  <options>       <dump>  <pass>
proc            /proc           proc    nodev,noexec,nosuid 0       0
# / was on /dev/sda1 during installation
UUID=d538b630-9997-463d-ac97-d4c06fbe1e87 /               ext4    data=writeback,noatime,nodiratime,errors=remount-ro 0       1
# /home was on /dev/sda2 during installation
UUID=b81776c9-2efc-4fd2-ae7b-a06e7e8662d7 /home           ext4    defaults,data=writeback,noatime,nodiratime        0       2
# swap was on /dev/sda3 during installation
UUID=de866c73-9960-4715-8d31-15180a0ac293 none            swap    sw              0       0
#SAMBA SHARE
//192.168.0.6/mybackups  /home/jason/backups  cifs  credentials=/etc/samba/user,noexec  0 0

Restart and enjoy.

If done correctly you should see a noticeable improvement in performance.