next up previous contents
Next: What's a Beowulf? Up: Introduction and Overview Previous: Introduction and Overview   Contents

Introduction

The Tao of Cluster Computing


In a little known book of ancient wisdom appears the following Koan:

The Devil finds work
For idle systems
Nature abhors a NoOp

The sages have argued about the meaning of this for centicycles, some contending that idle systems are easily turned to evil tasks, others arguing that whoever uses an idle system must be possessed of the Devil and should be smote with a sucker rod until purified.


I myself interpret "Devil" to be an obvious mistranslation of the word "Daemon". It is for this reason, my son, that I wish to place a simple daemon on your system so that Nature is satisfied, for it is clear that a NoOp is merely a Void waiting to be filled...


This is the true Tao.

-rgb




The following might be considered a ``recipe'' for a beowulf-class cluster supercomputer:

  1. Buy a pile of M$^2$COTS (Mass Market Commodity-Off-The-Shelf) PC's for ``nodes''. Details (graphics adapter or no, processor speed and family, amount of memory, UP or SMP, presence and size of disk) unimportant, as long as they ``work'' in the configuration purchased.

  2. Add a nice, cheap 100BT Network Interface Card (NIC) to each. Connect each NIC to nice, cheap 100BT switch to interconnect all nodes1.1.

  3. Add Linux and various ``ExtremeLinux/Beowulf'' packages to support distributed parallel computing; PVM, MPI, MOSIX, maybe more1.2.

  4. Blow your code away by running it in parallel...

Before I discuss the ``recipe'' further, there are some technical things that differentiate various sorts of cluster computing setups1.3. Many arrangements that more or less conform to the recipe above are not really beowulfs but are rather NOWs (network of workstations) or COWs (cluster of workstations) or POPs (Pile of PC's). Furthermore, some perfectly legitimate beowulfishly1.4 architected clusters are used to provide failover protection or high availability (e.g. webserver farms, transaction processing clusters) are not really beowulfs.

Clearly, we need to specify in some detail the answer to two questions, the first a FAQ on the beowulf list (sometimes answered when it isn't even asked). The first is ``What's a Beowulf''. The second is ``What is this book going to discuss''. So let's get to it.



Subsections
next up previous contents
Next: What's a Beowulf? Up: Introduction and Overview Previous: Introduction and Overview   Contents
Robert G. Brown 2004-05-24