DefinePK

Abstract

Scalability is a key concern for SMP based architecture in the current context. NUMA based architecture design seems to be a promising hope addressing the scalability. At the same time CC-NUMA based design architecture demands a deeper understanding and open vistas for key areas of improvement. With the approach of many core environments where memory is distributed among the different cores, it is challenging to design a thread scheduler along with proper data distribution across different nodes in a productive way. Our proposed research tries to investigate, evolve and analyze one of the key design issues for NUMA machine and proposes an innovative solution to address this key design issue under investigation in the current phase of our work. Our Algorithmic design proposed seems to be outperforming with respect to LibNUMA specifically.