Daily Archives: June 14, 2007

Btrfs, a new file system for Linux – from Oracle…

Chris Mason announced the new file system Btrfs at the kernel e-mail list. Btrfs is supposed to become a modern file system delivering all the features you might expect from modern file systems.

The announcement was featured at kerneltrap which also shows the most important features and discussions around Btrfs. There you also find the main features which are currently implemented (marked with a +) or which are supposed to be implemented in the future (marked with a -).

  • + Extent based file storage (2^64 max file size)
  • + Space efficient packing of small files
  • + Space efficient indexed directories
  • + Dynamic inode allocation
  • + Writable snapshots
  • + Subvolumes (separate internal filesystem roots)
  • - Object level mirroring and striping
  • + Checksums on data and metadata (multiple algorithms available)
  • - Strong integration with device mapper for multiple device support
  • - Online filesystem check
  • +Very fast offline filesystem check
  • - Efficient incremental backup and FS mirroring

Obviously, this reminds many people of ZFS – especially the snapshots are quite interesting, I think (think of backup solutions and versioned file systems here). But the file system is still in Alpha and in deep development. It has yet to show if it will be abpe to implement all planned features and if it will be able to become a new standard on Linux.

However, what I found most interesting was the fact that this new file system was developed at Oracle. Although Oracle attacked Red Hat with their unbreakable Linux the most kernel work is done by Red Hat in these days. The only bigger Oracle contribution I’m currently aware of is the Oracle Cluster File System 2, which is more interesting for larger deployments.

I wonder how many resources Oracle will put behind this development – and if their are other aims than providing a good file system. I’m still asking myself what their long term related to Red Hat is…

Anyway, if Btrfs would become stable, fast and usable quickly it could become the new standard for Linux file systems easily. At the moment everyone is looking at ZFS, but there are several problems why it is unlikely that it becomes part of Linux anytime soon. Btw., Dave Jones had a good point about the ZFS discussions: it is not all about the Licence. File systems in Linux are somewhat difficult and there are experienced reviewers missing to deal and properly integrate new file systems into the kernel.

However, since Oracle already brought another file system into the kernel and has lots of experiences in that field I hope that Btrfs matures well and becomes part of the mainline kernel soon. Just imagine the possibilities of a versioned file system in Linux within, say, 12 months. Ha, I like that.

KDE/Linux everywhere?

Today I saw that a drugstore where I bought a CD was using a KDE based system for managing there CDs and DVDs. This lead to the question how many hidden KDE installations – and Linux installations in general – are out there.

There are several big installations of KDE around these days – an examples is the Linux system of the city Munich, called LiMux. Other examples are some 80k school computers in Brazil and the German “Oberfinanzdirektion Hannover” with 12k KDE systems (basyskom was behind this one).

These examples are well known. However, since KDE is Open Source no one can track all installations, and several large installations might not be known yet.

The Müller drugstore chain

Today I discovered such an installation: when I bought a CD in a drugstore I realized that one of the two computers on the desk at the information was running KDE. It was heavily customized – no kmenu and a very simple browser window, for example – but kicker was obvious, and the Plastik window style is unique also. I asked the guy at the desk if that system was store wide or personal and he kindly informed me that that installation was used to look up and manage DVDs and other media.
The other computer btw. was some kind of very, very old DOS interface (you still see these often in commercial computer setups).

The drugstore was Müller, which is a drugstore chain with more than 400 shops in Germany and 60something in the rest of Europe. Every shop is running several computer stations (to get an idea: the shop I am talking about was spread over three floors!), but I’m not sure if they use the KDE system only for the media management or also for other purposes. But I think it it save to estimate that there are at least roughly 500 installations.

Of course, 500 installations alone will not change the world, but it is just one example of possibly many hidden installations – and I discovered it the first time today although I know this drugstore for more than four years. Also, as I mentioned, the other system on that desk was an old DOS system. Imagine they start upgrading these systems as well – it would be stupid upgrade these machines to yet another system so there is a good chance that these machines will be upgraded to KDE as well.

How many are we?

The question is now how many other installations are out there which no one realizes? And if you add all these up, what is the outcome? Of course Microsoft Windows is more spread (well, they do have a monopoly, after all), but Apple says itself has only 22 million installations – I wouldn’t be surprised if KDE+Gnome/Linux installations would have a bigger impact than that.

Unfortunately there are hardly any statistics. But half a year ago Mark Shuttleworth estimated 8 million (!) Ubuntu users. Also, Fedora reported almost 3 million installations of Fedora Core 6 in 26 weeks. Both numbers are impressive, but are old (in the Open Source world 6 month is old) and do not even include the corporate installations like SLED or RHEL, not to mention other Linux distributions like OpenSuse, Debian or Mandriva.

This again shows that the Linux community need to gather more statistical data: smolt and popcon are nice starts, but we need more. The best would be a cross distribution project to collect the data – at least among the community distributions, but that is unlikely anytime soon.

Still, even distribution specific data are very valuable. They show the real power behind the project, and they give the developers and the project leaders a stronger voice when these talk for example to hardware vendors about open drivers. Also, such things – done right – are good for advertising the distribution and Linux (or KDE or Gnome) in general:

Fedora Statistics - showing almost 3 million installations

Pretty nice, isn’t it? Thanks to the Fedora Statistics and Marketing people. We need more of such things!

Website as SVG?

Nicu Buculei programmed a web site entirely in SVG to check how it will be indexed by search machines.

The idea is as simple as interesting: since SVG is a W3C standard a web site programmed in SVG should be indexed by the search machines just like other web sites in for example HTML. However, no one really knows what the search machines do with SVG. Additionally many browsers do not treat SVG as well as they should.

The test site, called Nicu’s test website made with SVG, is entirely programmed in SVG – with no further additions. It is, btw., also an interesting example about the capabilities of SVG: menus, links, etc. But also shows that konqueror is missing a scroll bar in the SVG kpart…
Anyway, this web site will soon be indexed by the search machines – and after some days everyone can easily check if and how the page lands in the search engines. For this task there are several strings consisting of random characters embedded in each web page which makes it easy to search for them.

If you want to help Nicu with his task, you can simply link from somewhere to the SVG page – or write an article about it like I did ;)

This test btw. remind me of the Hommingberger Gepardenforelle, a (in Germany) famous search machine test done by the German IT magazine heise.de. Such tests are necessary from time to time – in case of the Gepardenforelle for example it turned out that Microsoft’s search machine discriminates Wikipedia, most likely on purpose.