DURIP: Equipment for Research on Heterogeneous Manycore Systems for Big Data Applications
Abstract
The objective of this proposal is to acquire suitable computing infrastructure to enable comprehensive investigations into design of high-performance and energy-efficient manycore systems for various big data applications. The acquired equipment will be used to create a new state-of-the-art laboratory at the School of Electrical Engineering and Computer Science (EECS), Washington State University (WSU). The new laboratory will significantly enhance the quality of research output of the ongoing heterogeneous datacenter-on-chip (HetDoC) project. Real-world applications including speech recognition, natural language processing, image and video analysis, networking, and cyber-security are extremely relevant for big data centers. However, the design of data centers is dominated by power, thermal, and physical constraints. On the contrary, emerging heterogeneous manycore processing platforms consisting of CPU and GPU cores along with memory controllers (MCs) have small footprints and offer power and area-efficient tradeoffs for running big data applications. Consequently, heterogeneous manycore computing platforms represent a powerful alternative to the data center oriented type of computing. The US Army Research Office (ARO) currently funds the PIs for the project titled ÒEnergy Efficient Heterogeneous Datacenter-on-Chip for Big Data ComputingÓ (Proposal No. 70970-CS). Successful execution of the research proposed in this grant is possible if we can analyze various big data applications including training large deep networks and big graph analytics using GPU servers. Moreover, the quality of the research output will be enhanced if we can prototype the HetDoC architecture using state-of-the-art FPGA boards. Hence, through this proposal the PIs plan to acquire two GPU servers and multiple FPGA boards. Specifically, the proposed computing infrastructure will help to achieve the following research goals: 1) Design a suitable network-on-chip (NoC) architecture as the interconnection backbone for the HetDoC that can handle both CPU and GPU communication requirements efficiently 2) Co-design approaches to trade-off accuracy and energy of deep inference on heterogeneous embedded systems 3) Evaluate power-thermal-performance (PTP) trade-offs for the proposed HetDoC by considering diverse big data applications We envision a HetDoC architecture targeting big data applications where the entire system (or a large part thereof) can be designed using a heterogeneous manycore-based single-chip architecture. Inspired by the successes of wireless NoC, we plan to explore design of heterogeneous (i.e., combination of CPUs and GPUs) systems combined with hybrid (i.e., combination of wired and wireless links) NoC architectures for big data computing. The uniqueness of, and the rationale for, the proposed research is that we introduce a new architecture for the next-generation big-data systems that are fast, cheap, energy-aware, yet widely accessible for running new algorithms with high societal impact. As an example, deep learning advances have been successfully leveraged to improve the performance of planning and decision-making under uncertainty algorithms. The key idea is to use deep learning techniques to automatically learn generalized representations for policies and value functions that form the core components of these algorithms. Army applications including real-time battle field planning, emergency response and logistics management will greatly benefit from the proposed HetDoC instantiation for deep learning techniques. Similarly, graph pattern matching over large-scale cyber networks forms the core computational process for cyber security and anomaly detection applications that are extremely important for Army. Our proposed HetDoC design methodology can be instantiated to provide a high-performance and energy-efficient computing platform for these security applications.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Mar 18, 2019
- Source ID
- W911NF1910162
Entities
People
- Jana Doppa
Organizations
- Army Contracting Command
- United States Army
- Washington State University