Blogs >> Technology >>
Linux Kernel - Features
Anyone following the rapid rate of change for the Linux Kernal version
2.6.32 I think will agree it is truly astounding. The work that is
going into this single piece of software is nothing short of
'incredible'. Not only to it's rate of change, but it's inclusion of
features, complexity and hardware support.
Before I get the the meat of my blog I'll summaries what is current and hip in the kernel at the minute:
1) Intel engineers are talking about changing the way pre-fetching works to increase performance by as much as 8%. In other words compiling not for space as previously but for speed.
2) Kroah-Hartman is looking at how to tag patches to increase work flow.
3) There is some talk about including legacy LIRC (Linux Infrared Remote Control) into the mainstream kernel. This was traditional provided by external drivers.
4) Those clever guys at DELL have proposed a new character device interface. Thus no more problems with device naming being inconsistent (e.g. on one machine a PCIA Ethernet card being named 'eth0' while another named 'eth1'.
5) Fast symbol resolution by some great work from Alan Jenkins. Thus a faster booting machine. And I thought it was fast already...
6) Para-visualization improvements for kernel-based virtual machines -KVM -Of which I won't go into details.
7) A fix for Linx Out Of Memory killer. This could take a blog in itself, but the fix basically changes where the Null page is stored in memory. Hence stopping a NULL pointer exception calling malicious software being called if referenced.
But all this great work seems to come at a price. It seems that the current 'big issue' facing the kernel is 'speed'. The performance of the kernel compared to previous 2.6 versions is down by 5-10% depending on what bench mark you use. It seems to be part of a trend that has a few people worried. To address this and help to get a handle on the issue the Kernel hackers have included in the /tools directory new benchmarking capabilities called 'pref bench'.
I guess the worry is that with Linux starting to go on to smarbooks, mobiles and netbooks nobody wants to see this issue have a negative impact on how Linux runs on small low powered devices.
Before I get the the meat of my blog I'll summaries what is current and hip in the kernel at the minute:
1) Intel engineers are talking about changing the way pre-fetching works to increase performance by as much as 8%. In other words compiling not for space as previously but for speed.
2) Kroah-Hartman is looking at how to tag patches to increase work flow.
3) There is some talk about including legacy LIRC (Linux Infrared Remote Control) into the mainstream kernel. This was traditional provided by external drivers.
4) Those clever guys at DELL have proposed a new character device interface. Thus no more problems with device naming being inconsistent (e.g. on one machine a PCIA Ethernet card being named 'eth0' while another named 'eth1'.
5) Fast symbol resolution by some great work from Alan Jenkins. Thus a faster booting machine. And I thought it was fast already...
6) Para-visualization improvements for kernel-based virtual machines -KVM -Of which I won't go into details.
7) A fix for Linx Out Of Memory killer. This could take a blog in itself, but the fix basically changes where the Null page is stored in memory. Hence stopping a NULL pointer exception calling malicious software being called if referenced.
But all this great work seems to come at a price. It seems that the current 'big issue' facing the kernel is 'speed'. The performance of the kernel compared to previous 2.6 versions is down by 5-10% depending on what bench mark you use. It seems to be part of a trend that has a few people worried. To address this and help to get a handle on the issue the Kernel hackers have included in the /tools directory new benchmarking capabilities called 'pref bench'.
I guess the worry is that with Linux starting to go on to smarbooks, mobiles and netbooks nobody wants to see this issue have a negative impact on how Linux runs on small low powered devices.
|
|
