SGER: Dynamic Partitioned Global Address Spaces for Future Large Scale Systems

Georgia Tech Research Corporation, Atlanta GA

Investigators

Abstract

With the continued explosive growth in electronic information, both business enterprises and high performance computing systems are being structured as large ensembles of interconnected modular computing/storage modules or blades. These systems form the next generation data centers and supercomputers and a major impediment to their successful wide-spread deployment is the growing capital cost as well as power and cooling costs. A dominant contributor to these costs is the memory subsystem that can exceed the costs of the processing core and consume as much as 25%-40% of the energy in high end configurations. Currently, the architecture of data centers limits how physical memory in data centers can be shared by different programs in the system. Consequently, systems must be designed to accommodate the worst case peak memory demand and is therefore unduly expensive both in terms of capital cost as well as power. The proposed research explores a novel solution for the reduction of memory system related costs by enabling memory to be shared across blades in a dynamic, demand-driven, fashion. The key idea is the tight integration of memory subsystem with the network subsystem. The result is a significant increase is memory efficiency and an attendant drop in the total memory requirements of a data center with little to no performance penalty. This approach is made feasible by exploiting recent advances in chip integration where the memory controllers and high performance communication interfaces are integrated on the same die. Thus cost and power can be more effectively managed than previously feasible by tracking actual memory demand rather than statically provisioning for worst case memory demand. To maintain its leadership in computing, U.S. industry requires both technical advances such as those proposed here, new companies that leverage these advances, and new employees with the skill sets needed for rapid technology and product development. The proposed work offers suitable technical elements to offer ways to engage students in Computer Science and Engineering topics, to arrest the ongoing decline in the U.S. enrollment in CS and CmpE, to create new employment opportunities and to leverage the creativity and entrepreneurial nature of the U.S. student population. Existing relationships with technology companies such as IBM, HP, and Intel through an existing NSF I/UCRCat Georgia Tech will provide pathways to influence the system community.