LinuxConference Europe 2007 1/X
After our first night in Cambridge with Tab and Gildas where we expierenced some traditional English lack of culture , I woke up this morning ready for the first LinuxConference Europe , the schedule looks good I hope the conference lives up to my expectations.
So the first stop of the day was in Johnatan Corbets Kernel Report talk.
- On Release Cycles
The curent release schedule of the kernel is different from what it used to be 2-3 years ago. But now we should be able to predict when new kernels will be released.. this looks like a good thing -
On Kernel scalability, remember the time when people told you that Linux couldn't handle 2-4-8 or 16 CPU's ? Well Linux handles about 1024 CPU's easy. Things are starting to get rough when you want to ma
nage 4096 CPU's. About 20% of the systems memory then is used to actuall manage those CPU's but hey I guess that by the time I got 4096 CPU's in my they will have fixed that problem ;)
scalability works both ways up, but also down. The embedded linux users really should start to contribute more to the kernel , not just fork and forget but contribute their changes back. -
On Filesystems, there is a lot of stuff going on the kernel regaring filesystems. Ext4 is peeking around the corner, btrfs is coming , LogFS is comming and most of those filesystems are being discussed
at the conference. Johnatan mentions that one of things we are struggling with is fsck. Big disks mean long fscks, one theory is to split a filesystem in multiple parts, and monitor which parts of the fils
ystem have bveen modified so you only need to fsck those. The basic idea of this concept is pretty simple but the implementation is a bit more difficult, think about what should happen when a file on such
a part grows beyond the available free space on that part, how do you manage that ? However this still need to be implemented. - On Virtualisation, lots of stuff going on there. Xen, Containers, KVM , lguest etc are all growing and Rusty even claims that these days there is more to virtualisation than collecting money from VC's
New features in the kernel these days seem to have to be "Hellwig Approved" :)
As on the GPLv3 question Jonathan mentions that it will probably never happen in the kernel, one of the biggest reason is that you can't possibly trace back everybody who once contributed to the kernel , l
et alone have them all agree on the chance of license.
Next step was the OS Circular talk by Kuniyasu Suzaki
OS Circular is a framework of Internet Disk Image Distributor for anonymous OSes, it bases lots of stuf on Fuse and supports different Virtualisation techniques, pretty interresting.
It reminds me of the original Xenoservers project. If you are looking in this direction also have a look at Floz (Free Live OS Zoo)
Too bad Stephen Heminger wasn't at the conference in time so we had to head into Jeff Dike's talk ..
As Alasdair told me he still had a free slot on monday afternoon which Stephen could fill up if he showed up and I just ran into him in the hallway I'll be trying to catch his talk later.
One of the talks I had been looking forward to was the Ganeti talk, I was pretty dissapointed by the lack of depth in this talk and the general approach, honestly I don't care on how many machines you are
deploying it. I wan't to know how it works, what you are using underneath etc. I have to download the tarbal and look into it , but it seems they are pretty old versions of Xen and DRBD .
If in the next decade I find some spare time .. I`ll be talking about my experiences with it here ;)
More later ..