New Computer Architecture Method Aims to Break the Big Data Bottleneck
As the size of data sets continues to grow in the era of big data, researchers at Illinois Institute of Technology are working in stride to develop more efficient methods to process this data.
Rujia Wang and Kyle Hale, assistant professors of computer science, and Xian-He Sun, distinguished professor of computer science, have received a grant from the National Science Foundation to build a new computer architecture that will reduce the energy and time needed to process big data sets.
“In the computer systems we are now using, you have a CPU [central processing unit] and memory to store data,” Wang explains. “When you want to process information, that data must be moved from memory to the CPU, which is OK when you don’t have a lot of data. But if you are working in machine learning, data mining, or something specific like genome sequencing, that’s a lot of data movement.”
Bottlenecks occur as there is limited space where data can travel from memory to the CPU, which creates slower processing times and a greater use of energy.
This proposal is to develop a method to start preliminary processing of the data in the memory before moving it to the CPU. This should reduce the amount of data that needs to migrate to the CPU and will make processing more efficient in the CPU.
To accomplish this, Wang says, new software and hardware must be developed and be able to work with their existing counterparts. The Illinois Tech team of Wang, Hale, and Sun bring their collected experience and expertise in hardware, software, and high-performance computing systems together, forming a strong team to solve these memory system issues.
“The biggest challenge we’ll face is developing a full stack for hardware that is not really ready for prime time,” Hale says. “There is really only one commercial product in the works for near-data processing, and it is not yet on the market. This makes things challenging and interesting for us on the software side because we either have to use simulated hardware or develop clever ways to approximate the future with hardware we already have.”
Wang says this new architecture would have an immediate impact in the high-performance computing and supercomputing fields by increasing data processing speed and efficiency.
“This project is exciting because it is fundamental, and it’s potential impact is unlimited,” Sun says.
The opinions, findings, and conclusions or recommendations expressed are those of the researcher(s) and do not necessarily reflect the views of the National Science Foundation.
Photo: Assistant Professor of Computer Science Rujia Wang