Clusters mean a set of computers, working together to solve a problem, which individual computers may take a long time to process. Problems commonly solved by clusters are mainly arithmetic problems like finding out very large prime numbers. Linux is probably the most used cluster operating system in the world today. Organisations like universities can achieve supercomputing speeds using such clusters, without actually having to buy an expensive supercomputer.
A widely popular clustering technology in use today is Beowulf Clusters. They are very powerful, but special application coding has to be done to harness their true power. They use the clustering libraries PVM and MPI. This is very difficult for the common user. But recently, technologies have come up which let people take advantage of the technology of clustering without having to do application coding.
A popular distribution which lets users harness the power of clustering is Parallel Knoppix. It is a modified form of Knoppix, which itself is derived from Debian. It uses the LAM-MPI and/or MPICH implementations of MPI. Since it is derived from Knoppix, it runs off a live CD. The hardware detection ability is quite impressive though the security could be beefed up a little bit.
OpenMosix is a technology that adds clustering abilities to the linux kernel. The main advantage of OpenMosix is that applications don't need to know whether they are running on a cluster or on a local machine. Also no special programming is required to take advantage of the OpenMosix technology.
The concept of clustering makes the user feel as if he is running his applications on a large Symmetric Multiprocessor (SMP) System, whereas all that he is actually using is a bunch of slow or medium speed linux machines. The bottleneck that occurs here is that in an SMP system, the task is divided locally into different processors, but in a linux cluster, the task is divided into a large number of machines over a LAN. If the speed of the LAN is slow, it results in a significant drop in efficiency. Thus fast LAN technologies like Gigabit Ethernet must be used when clusters are in consideration.
To build a linux cluster, first you need a standard hardware configuration in tune with the Linux Router Project (LRP). If you don't want to use a modified kernel specifically built for clustering, I would suggest go for Kernel 2.6.5-1.358, distribution Fedora Core 2. Also for your clustering needs a common firewall setup, using IPTables would do. Though be careful while coding the ALLOW and BLOCK scripts.
Of course, all this task can be reduced to a minimum by using cluster specific distributions like Parallel Knoppix described above. You can also use FAI, which is an automated system to install a Debian GNU/Linux operating system on a PC cluster.
References :-
- http://www.ibm.com
- http://www.ram.org
- http://lcic.org
This article was originally published by CyberArmy.net in the CyberArmy Library.
|
|