How to make supercomputer in home ?

We can easily make a supercomputer in home very easily . Are you thinking , Is this possible ? then yes , we can easily get the performance of the order of a supercomputer by just connecting our own PC’s with each other and after that by installing some open source software and a small configuration according to our need that’s all. Now you have your own supercomputer. This computer is known as cluster computer.

Cluster computer means connecting commodity PC’s together by a high speed network to form a single stand alone computer. Cluster computer is basically of three types

(1) Failover cluster (High availability HA cluster)

(2) Loadbalancing cluster

(3) High balancing cluster (HPC)

(1) Failover cluster is used for providing high availability of the services, when one server is fell down then other system can take all the work of the failed system. Hence provide high availability. (e.g. opensolaris HA cluster )

(2) Loadbalancing cluster is used, when there is a lot of work is put on the main system, then this computer takes help of its connected computers to execute instruction for it. Hence load is balanced between all the connected computers . this load is balanced by just some modification in the kernel of the main system , we have to add patch in the kernel for load balancing and some configuration on other node computers , that’s all . (e.g. openmosix cluster).

(3) High performance cluster computer is almost same load balancing cluster in the sense that it also distributes its tasks between other nodes, but in HPC there is no changes in the kernel of the main system. We have to connect all the computer, then we install a MPI standards (e.g. MPICH2 , open MPI , LAM/MPI ) then some configuration , that’s all . After that if you want to use this cluster to run an application , then that application should also have a parallel implementation possible and you have to write that application using MPI standards.(e.g. boewulf cluster).

No Comments

This post is password protected. Enter the password to view any comments.

Post a Comment