NSF Award
1519508
The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase I project is to enable the scaling of data centers in a very cost effective way. Global information technology business is reaching $3.7 Trillion, and data centers are an increasingly important cornerstone of future global information technology infrastructure. By providing timely solutions to improve the cost effectiveness of data centers, this project will contribute to accelerating the penetration of data center based cloud services into a wider spectrum of applications and hence generate considerable technical and societal impact. In addition, as the project disruptively relaxes the traditional distinct boundary among computation, storage, and networking, its success will influence the entire industry to fundamentally re-think and re-innovate data center design and implementation for future information technology industry advancement.<br/><br/>This Small Business Innovation Research (SBIR) Phase I project will develop a working demonstration of a disruptive product that can significantly improve the cost effectiveness of future data center infrastructure. With inevitable slowing-down of Moore's Law and rapid growth of data volume, the conventional data center design practice demands formidable capital investment, which will seriously undermine the cost effectiveness of future data centers. Aiming to fundamentally address the data center cost effectiveness challenge, this project will develop a working demonstration of a software/hardware integrated product that can cohesively handle data storage and processing tasks in data centers at very low cost. The key intellectual merit is to fundamentally re-think the realization and coordination among computing, storage and networking. The product can outperform CPUs by at least an order of magnitude for critical data processing tasks in data centers and meanwhile significantly improve memory/storage hierarchy efficiency. The software/hardware integrated products installed across multiple server nodes can further form a virtualized platform, with which inter-node data transfer and computation coordination can be realized very efficiently.
