The second day of LinuxCon started out with some keynotes on Cloud Platforms, but since the subject didn’t fancy me, I spent the morning at the KVM forum where I got the “weather report” for the community. KVM looks like a very strong community with lots of achievements and many plans for the future. Though I found out I was a total outsider to KVM, I did learn about the “Bit KVM Lock” that is a serious issue or the scalability of the system.
The rest of the morning, I visited the stands back at LinuxCon. Most of the time was at Intel’s Tizen booth. I did know about Tizen being a Linux Foundation backed project as an alternative to Android, but that’s about it. I asked the Tizen guys _a lot_ of questions. I found out that the architecture is very similar to Android’s but it relies more on features already available in Linux (such as SELinux) rather than reimplement new things (like Android does with the Dalvik virtual machine). I learned that they have their own application development environment (also Eclipse based) but rather that make their developers write Java, they make them write HTML5. This made me think if apps made for Tizen could be easily ported to FireoxOS and vice versa (we shall see). They also have a native app environment where you can write C based applications. And I was curious how a HTML5 app would communicate with a native app. The answer was: WebSockets. Tizen is to be released next year, and Samsung, Intel and other companies have invested a lot in this project.
Seeing the Tizen presence at the conference, made me think about something: there was absolutely no Android/Google at LinuxCon. Although Google’s Android (and to some extent, ChromeOS) was praised during some presentations for bringing Linux to the everyday user and making it so embedded in the world, I was surprised that Google or someone in the Android community had sent representatives to talk about invocations of the project. Is it maybe because “Android has won”, there is no need to invest in talking about it anymore? That seems a mistake.
One the the highlights of the day was the Kernel Developers Panel (video available on video.linux.com). Hosted by LWN’s Jon Corbet, its main guest was Kernel maintainer Greg Kroah-Hartman. Greg talked about how hard it is to maintain kernel subsystem because it’s difficult to find and keep committed maintainers. He also pointed out the fact that, apart from tweeking some features, the core of the Linux Kernel is pretty much stable and almost all development is being done on device drivers. And the need for them is always present.
After the Developers Panel, I went to a presentation hosted by the same Jon Corbet from Linux Weekly News: The Kernel Report. It was probably the most interesting presentations of the conference. He started with some statistics and observation on those statistics. First of all, the time between releases has shrunk (but he noted that is is a limit on how small a release cycle can be, and we’re getting there) and the number of contributors is on the rise. But the number of independent/unaffiliated/hobbyist contributors has gone down and the number of company backed contributors has gone up. Still, the need for kernel developers is still high. This and along with other statistics will, probably, be available online for the yearly report. On interesting statistic that he made for the origin of the patches sent to the Linux Kernel was based on the timestamps of the emails (because the domain name of the email doesn’t help a lot). The proportion was : Europe (I’m assuming it also includes Africa, though not included in the presentation) ~40%, the Americas ~30%, Asia ~20%.
Looking to the future, the Kernel Report listed some interesting technologies that should be focused on in the next years. It included things from mobile to data centers, from data storage to networking and security. Among the things mentioned was, for example the need to think about using huge pages by default in the system’s memory management. Seeing how memory resource needs have grown, the page size introduced 20 years ago might not scale anymore. And he pointed of this trend in OSv, an operating system built for the cloud (virtualization) that only uses huge pages. The system scheduler is another components that needs tweeking in order to provide better power management, and mentioned the idea of the tickless kernel. Improving filesystems and SSD support is a very important topic, mainly in the data center world. Multipath TCP is something needed in the Internet, but the Linux kernel is falling behind in implementing it (especially since Apple announced that it was implemented in its iOS). The new nftables was mentioned. It is a new firewall system that is to replace the 4 current tables (iptables, ip6tables, arptables and eptables) into a common system.
At the end of the day, googlers from the Google Open Source Program gathered participants in a room for a discussion of the Google Summer of Code program. They wanted to get opinions from past mentors and students and gather feedback on how get open source communities more involved in this project buy offering more mentors.
To be continued.