Memory Management And Optimization Using Distributed Shared Memory Systems For High Performance Computing Clusters
About Book
Advances in microprocessor and network technology have recently made it possible to create affordable supercomputers using
clustersof cheaper processors (Acher, Hellwagner et.al 1996). However the operating systems involved still deal with the hardware as a collection of loosely connected processors each with local memory. In the area of cluster memory management, there presents a great challenge to allow existing supercomputer software to run on cluster hardware. This could be achieved by modifying the memory management services to support extra facilities such as inter-computer memory messaging.
This problem of changing the memory management requirements of all the diverse applications which might run on clusters has resulted in very few becoming available for clustered supercomputers (Abrams, Doraswamy, Mathur 1992). Instead of changing the individual memory management of each application, it has been proposed to modify operating system to provide clusterwide memory management, which will convert distributed memory into transparent global memory for any existing supercomputer application.
Read More
