Interview: Theodore Ts'o
Theodore Ts'o will give a talk about Ext4 at FOSDEM 2009.
Could you briefly introduce yourself?
I have been a Linux kernel developer since the very early days (August, 1991, version 0.09), and am the first North American Linux Kernel Developer. I am the ext4 filesystem maintainer, and primary author and maintainer of the e2fsprogs utilities for ext2/3/4 filesystems. I am a member of IBM's Linux Technology Center, and I am currently on a two-year assignment with the Linux Foundation where I am serving as the Chief Technology Officer.
What will your talk be about, exactly?
My talk will cover the new features in the ext4 filesystem, why it is better than ext3 and how it compares to other filesystems. I will also talk about the history of its development.
What do you hope to accomplish by giving this talk? What do you expect?
I hope to encourage people to try out ext4 and upgrade their ext2/3 filesystems to ext4.
What new functions will be added to ext4 in the near future?
We are still working on on-line resize support, and the 64-bit block number support for filesystems larger than 16 TB is still undergoing testing.
Is ext4 ready for production use as a replacement for ext3?
I have been using it as my laptop filesystem since July. We're still getting some bugs reported, so I wouldn't recommend it for critical servers just yet, but for personal laptops and especially for people who are willing to keep up with the latest kernels it is something that people should definitely try.
Editor's note: Since the interview, a stable ext4 was released in kernel v2.6.28.
Is ext4 only useful for users who have more than a few terabytes of disk space?
No, ext4 has many features that make it useful for personal systems, such as much faster fsck times, much better performance, and better fragmentation resistance.
When will GRUB support booting a kernel from a ext4 partition?
There are patches for GRUB and GRUB2 available for ext4 today. So it is really a matter of when a particular distribution has full support for ext4.
How big of a role will ext4 play within Linux, knowing already that BTRFS will be the next generation filesystem for Linux?
Ext4 was designed to be something that could replace ext3; it isn't necessary to reformat a pre-existing ext2 or ext3 filesystem to use the ext4 filesystem driver. This makes it highly attractive for users who want an easy migration path. In addition, while I do believe that btrfs will a very good filesystem in the long term, and one that will likely supplant the ext2/3/4 filesystems, it takes a long time for filesystems to mature. (For example, many system administrators did not trust ZFS for production use on critical servers for 5-6 years after it was first introduced.) As of this writing the on-disk format for btrfs is still being finalized, and btrfs does not have yet a rich set of filesystem support utilities such as is provided by e2fsprogs' e2fsck and debugfs.
One of the critical advantages of ext4 is that it reuses a lot of very well tested code from the ext2 and ext4 filesystems. This conservative design means that we have been able to stablize the filesystem much more quickly, and we have been able to leverage utilities originally designed for ext2 and ext3. In addition, and very importantly, there are many people all over the world who are familiar with the implementation details of the ext2 and ext3 filesystems, which will make it easier to find skilled engineers to provide support for ext4 --- both in the community and in the commercial enterprise distribution market.
Now that you are the new CTO of the Linux Foundation, how does this change your focus?
The CTO is a new title for me at the Linux Foundation but the sort of things that I have been working on hasn't changed much from last year. I am still working hard on ext4, making the Linux Standard Base better and to get more distribution and ISV applications certified to the LSB, and helping to organize various Linux Foundation events such as the Kernel Summit, Linux Con, and the Japanese Linux Symposium.
This interview is licensed under a Creative Commons Attribution 2.0 Belgium License.