Tim Burke, vice president, Linux Engineering at Red Hat, shares his thoughts about the new Linux kernel 2.6.38, how Red Hat is going to push it to their customers and what is the future of ext4 amid the arrival of BTRFS in this exclusive interview. Read on...
Muktware: How does the new kernel improve speed and performance?
Tim Burke: Any specific upstream of the kernel can impact lots of different work loads in unique ways. If you look at the coverage of the key features in 2.6.38, some of the more high profile additions were given an excellent overview in both LWN and at the kernel newbies page at:
To add some details that they did not mention in either summary, we also have a lot of new code in the NFS client in 2.6.38 that will be the building blocks for parallel NFS (a way to make clustered NFS servers much higher performance).
Muktware: How will Red Hat push the update to its customers?
Tim Burke: Depending on the maturity of the feature, the timing for release in a Red Hat Enterprise Linux release will differ.
Some of the features in 2.6.38 are already shipping to Red Hat Enterprise Linux customers, like the transparent huge page support and btrfs which is in technical preview in Red Hat Enterprise Linux 6.
Red Hat engineers work actively to develop these features in the upstream kernel, work both with the Fedora community and others to integrate them for early adopters and then evaluate when they are robust enough for enterprise users.
Muktware: As a major contributor, what is Red Hat's opinion about the way Linux kernel is shaping up?
Tim Burke: Red Hat is clearly very passionate about contributions to the Linux kernel and employs a large number of people who contribute to it on a daily basis. We are quite happy with the progress that the upstream kernel has made and its continued record of excellence in adding support for new hardware, higher scalability and other enterprise features that our customers need!
Muktware: What are the plans to replace ext4 with BTRFS?
Tim Burke: File systems are very long lived in the field since users tend to maintain them for years and years. In many cases, enterprise users have external storage so the storage will survive several generations of servers. That means that ext4 is not going anywhere any time soon!
That said, Red Hat is an active contributor to the btrfs project and we are actively working on getting some of the key inhibitors for wide btrfs deployment finished with the upstream community. To enable our users to help in this effort, we ship btrfs in Red Hat Enterprise Linux 6 as a "technology preview" item which means they can test btrfs from a normal Red Hat Enterprise Linux 6 machine. Technology preview items are actively refreshed against the upstream source, but are not technologies that we support for production use.
We hope to see btrfs reach a level of stability and features in the coming year or two that will allow us to bring it to our users in a supported way.