The Plain Vanilla Cluster
What is it ?
The Plain Vanilla Cluster is a simple means of installing and configuring a small bare-bones Beowulf cluster using a kickstart DVD made from the CentOS 6.4 DVD ISO disk image and a scripted configuration package. The Plain Vanilla Cluster is Licensed under The GNU General Public License Version 3.
Building a Beowulf cluster from scratch is actually rather easy, but can be time consuming. In the past there were quite a few very nice packaged cluster distributions , which have basically gone away, and about all that is left is some of the more complex, and highly constraining cluster distributions that really don't seem to fit the need for a small, simple, efficient Beowulf cluster package that is reasonably easy to stand up in a short period of time.
What do I need to do this ?
You need some computers (PC's) to build the cluster. A good recommendation as a starting point is somewhere between three and six computers will make a nice cluster. It is easier if all of the computers are the same... identical if possible, but otherwise use what you have.
You will need a network switch or router with a built-in switch to connect your cluster machines to. An inexpensive home firewall router of the type that many people use for DSL ought to work just fine. Most computers have Gigabit Ethernet, which will be adequate for large clusters.
Your computers need to have DVD drives to install the operating system from DVD disk.
You need to visit the nice folks at CentOS and download the CentOS 6.4 DVD: CentOS-6.4-x86_64-bin-DVD-1.iso
You will need root access to a computer running Linux which has a DVD R/W drive so that you can create the kickstart DVD.
You need to download two packages, possibly three, from this site:
1) pvc-kickstart.tgz contains the files and instructions for creating a kickstart DVD that will automatically install CentOS 6.4 onto the computers used in your cluster. Be sure to read the documentation completely so that you don't accidentally lose data.
2) configs.tgz contains the script, files and information you need to configure the computers used in the cluster. Be sure to read the documentation before attempting to use these tools.
3) ptools-1.3.0.tgz is a set of simple, lightweight, and fast command line tools for sending shell commands and copying files among many computers at the same time. This package is not required, but is recommended. It makes day-to-day work on a Beowulf cluster a lot easier.
How do I do this ?
1) READ THE DOCUMENTS FIRST
2) Use the files in pvc-kickstart.tgz in combination with the CentOS installation DVD to create the kickstart DVD.
Click HERE to read the instructions for making the kickstart DVD.
3) Install the kickstart distribution onto all of the computers to be used in the cluster. Make sure that the computers have only ONE hard drive installed. Make sure the computer is configured to boot from the DVD drive. Just put the disk in the dvd drive and boot the computer.
4) After the computers have installed prepare to use the configuration package to configure all of the cluster computers. Click HERE to read the instructions for using the configuration package.
5) Click HERE to read a brief description of the changes that setup.sh makes to the operating system configuration.
6) The basic cluster should be ready to go at this point. Move in , install your favorite tools and applications, and have fun.
Does it scale ?
Yes, the PVC cluster can scale to a large number of nodes. If you need to build a bigger cluster, just build it bigger.
If you are using PVC, I'd like to hear from you. If you found a bug, I'd appreciate hearing from you.