4.2 Linux Kernels and Linux Distributions

		Once Linux is installed on a single node, there are options for loading the rest of the system. You could follow the preceding procedure for all of the cluster nodes, but that would be time consuming and error prone if you have more than a very small number of nodes. Better ways to install Linux on the rest of your cluster are discussed in Chapter 6.

		4.9 Keeping Up with Linux

		At this stage you should have a basic understanding of what Linux is, what it can do, and how to set up a basic system. While we have tried to give more than just a cursory overview, it is impossible to capture all the details in a single chapter and we have glossed over some topics. You should keep in mind that installing Linux can be a time consuming process. Even though the installation process is made much easier today by menu-driven installation programs, you may well make mistakes the first few times and have to start over. After you've done this once or twice, future attempts will be far easier.

		New releases Even though we have constantly referred to Linux as a static entity, it is really a moving target. Linux is constantly evolving as new work is always being done on the kernel. No central company or group controls the development of all the software that is typically run on Linux. Many independent programmers scattered across the world, working on their own and sometimes in groups, develop and maintain most of the software. The many people and organizations using the software submit bug reports when they run into problems, and most of the time the authors release a bug fix in a few days, or the bug finders submit their own patches. Since the source code to all of the software is freely available, users can fix problems quickly without necessarily having to rely on the author. Consequently, new versions of software are released on a monthly, and even weekly, basis. It is easy to feel overwhelmed by this rapid pace of change. We recommend that you only upgrade software as needed, otherwise you will be playing a constant game of catch up.

		The free availability of Linux source code has made it possible to advance Beowulf research. Scientists and hackers alike have been able to modify and extend Linux source code to enhance network performance and add performance monitoring subsystems. Without this ability to customize the kernel, it would not have been possible to attain the performance demonstrated by the very first Beowulf, which used dual networks that logically appeared as one to increase network bandwidth. Those modifications, along with more recent software that is necessary to build a