$9.2M grant to UChicago computer scientists will improve graph analytics

With a $9.2 million grant from the Intelligence Advanced Research Projects Activity (IARPA), Prof. Andrew A. Chien will lead a team of computer science researchers at the University of Chicago in building UpDown Systema, a new approach that could speed up graph analysis a hundredfold.

Graph analysis is at the heart of some of today’s most exciting computing applications in science and technology. Organizing data into graphs—large networks of people, molecules, or locations connected by their interactions and relationships—can provide powerful insights into e-commerce, scientific discovery, social networking, recommendation and search engines, and fraud or anomaly detection.

However, modern computing architectures are not designed for graphs and struggle with efficiency and scalability.

An IARPA grant from the U.S. Intelligence Community’s research arm will fund the development of the UpDown system to accelerate graph analysis. This effort will reinvent computer architecture by dramatically increasing the efficiency and scalability of graph computing. This scale will be required to effectively analyze the world’s largest graphs from social media, financial transactions, or IoT device networks with billions or trillions of vertices and edges.

Also Read :  ‘I hurt my baby,’ capital murder jurors hear Mobile man tell police officer

“Efficient and scalable computation using massive graph structures is the signal computing challenge for the next several decades,” said Chien, who is the William Eckhardt Distinguished Service Professor in the Department of Computer Science and a senior computer scientist at Argonne National Laboratory. “Our invented UpDown architecture has new capabilities for both efficiently encoding information and intelligently moving it around the machine, both of which are essential for faster graph computing.”


Chien will lead a team of computer science researchers at UChicago, including Henry Hoffman, Yangjing Li and Michael Mair; the team also includes graph computing experts from Purdue University and Tactical Computing Laboratories, who will develop and test the UpDown hardware.

Changing priorities

For decades, high-performance computer architectures were optimized to perform linear algebra and other dense operations, and improvements focused on single-value transformations (known as floating point) rather than the complex, intelligent data movement required by complex data structures. Dealing with large and sparse graphs changes priorities: computers perform fewer operations, but must intelligently interpret and move a much larger volume of data.

Also Read :  Affinity 2 Photo, Designer and Publisher apps launch

Thus, the UpDown architecture provides flexible graph representation and programmable intelligence to move it around the system. The UpDown accelerator sits between the CPU and memory, giving an application the ability to create “software-defined hardware” that customizes how data is encoded, interpreted, and moved around the node. Programmers can write software to improve the performance of specific applications or use machine learning to optimize the flow of data through a system.

“The UpDown architecture supports programming paradigms that weren’t viable before,” Hoffman said. “For example, UpDown can transform data representations on the fly where previous work would have forced programmers to create a single compromise representation. We will use automated methods based on our previous work on self-awareness to take advantage of this opportunity by choosing the best representation at each point. The goal is to intelligently obtain maximum value from this new architectural capability.

Also Read :  Vertiv Introduces Water-Efficient Liquid Cooling Solution for High-Density Data Centres in Europe, the Middle East and Africa (EMEA)

In preliminary evaluations with scientists and applications at Argonne National Laboratory, the UpDown team demonstrated a 100x performance increase in several graph computing benchmarks. These improvements not only speed up graph analysis tasks, but would also make them more energy efficient.

In the next phase of the project, the team will build on the radical design of the UpDown system, exploring new applications as well as scaling the architecture to tens of thousands of nodes.

“UChicago’s selection reflects the institution’s leadership in large-scale systems, computer architecture and machine learning, as well as the collaborative synergies between these areas at UChicago and Argonne,” Chien said.

The UpDown project was funded as part of the IARPA AGILE (Advanced Graphic Intelligence Logical Computing Environment) program. Additional members of the UpDown team include David F. Gleich of Purdue University and Dave Donofrio and John Leidel of Tactical Computing Laboratories.

-Adapted from an article first published by the Department of Computer Science

Source

Leave a Reply

Your email address will not be published.

Related Articles

Back to top button