NewsOctober 21, 2013
Systems Researchers Enjoy High Success Rate in Grants
In light of the current restraints on federal funding, the Systems Research Group in the Department of Computer Science scored something of a coup this fall. The group, with five core faculty members, was awarded four National Science Foundation (NSF) grants, totaling nearly $1.6 million. “Our success rate was 50 percent, which is pretty remarkable,” said Michael Scott, Professor of Computer Science, who is PI or co-PI for two of the awards.
A major focus of the Systems Research Group is parallel computing, in which many calculations are carried out simultaneously. A proliferation of multicore processors – ranging from supercomputers to desktops and smart phones, incorporating GPUs and other accelerators – has enabled parallel computing on an unprecedented scale. But this has also created a host of challenges, including uneven energy efficiency, interrupted access and vulnerability to malicious attacks.
The projects funded by the new grants tackle these issues, and have at least a peripheral connection to Big Data research, Scott noted, because they involve “managing volumes of data that far exceed what we have managed in the past.”
- Scott and co-PI Kai Shen, Associate Professor of Computer Science, have been awarded a three-year, $455,255 grant to better manage graphics processing units and other accelerator devices, which have become “increasingly ubiquitous in computer systems.” Because these devices are specialized, they can perform certain tasks much more quickly and efficiently than a central processing unit (CPU). But because they are programmable, unmanaged access can open the door to significant unfairness or even malicious attacks. The goal of the grant, Scott says, is to come up with safe, efficient techniques to minimize these problems in standard operating systems, across all vendor products. “A lot of Big Data has to do with pattern matching, with mining for certain kinds of information in very large volumes,” Scott noted. “A lot of that can be accelerated by running it on GPUs or other special hardware. The techniques we are developing will be useful in those kinds of applications.”
- Sandhya Dwarkadas, Professor of Computer Science, is PI, and Scott is co-PI of a three-year, $449,911 grant aimed at making “big data” computing more readily available for projects involving large sets of irregularly structured data. The goal is to identify a “richer infrastructure” of design patterns than recent software packages offer, especially for applications that use trees and graphs for data representation and access. The project will also develop efficient run-time software to support the new designs. This would enable “advances in scientific computing in general and computational biology in particular,” Dwarkadas notes in the award abstract. “We are particularly excited about the opportunity to use the Blue Gene Q (supercomputer) on this,” Scott added. “It is a good machine for graph computations because it has a lot of cross-sectional bandwidth and low latency for communication.”
- Dwarkadas is PI and Shen is co-PI of a three-year, $433,673 grant that addresses the energy efficiency and dependability of multicore systems-on-a-chip that have “permeated every segment of the digital market, from servers and supercomputers to desktops and smart phone mobile devices.” “A lot of the issues with power management and protection are ones we used to have in servers, but now are affecting the smaller devices,” Shen said. Multicore processors allow small devices to perform unprecedented tasks. The tradeoff, however, is uneven power usage that can quickly deplete batteries. Part of the challenge with small devices, he added, is that their power usage can fluctuate widely depending on which programs are being used. One goal of this grant is to develop power containers that will “identify and mitigate either malicious or unintentional power anomalies that result in unusually high power consumption, in both high-end servers and multicore and multi-accelerator smart phones,” Dwarkadas and Shen note in the award abstract.
- Chen Ding, Professor of Computer Science, is PI for a three-year, $253,357 grant to adapt Ruby, one of the increasingly popular, easy to use dynamic scripting languages, for parallel computing. “Increasingly, non-computer science people write programs. They want to spend their time doing physics, doing science, doing medicine -- instead of writing code,” Ding noted. The project will augment Ruby with constructs to enable both parallel and sequential programming. Students will be involved in testing the augmented Ruby, Ding said, so that “we should have a prototype that’s fairly mature, that has a lot of user experience and documentation and will be ready to be used by the community” when released as open source code at the end of three years. One of Ding’s major goals is to use the augmented Ruby as a “gentle introduction” to parallel programming for students across many disciplines and levels of expertise.
Two other NSF grants were received this fall by Computer Science faculty members. Henry Kautz, chair of the department, received a three-year, $481,939 grant to further develop his studies of online tweets as a way to identify, track and analyze disease outbreaks. Daniel Stefankovic, Associate Professor of Computer Science, received a three-year, $399,703 grant to identify sampling problems with efficient algorithms.
Computer Science faculty members have been lead investigators on a total of 20 NSF grants awarded since July 2010, totaling more than $8.8 million.