Database Systems Expert Joy Arulraj Joins School of Computer Science

Database systems is one of the most wide-reaching fields in computer science, influencing everything from healthcare to government. For new Assistant Professor Joy Arulraj, database systems allow him to make an impact in as many research areas as possible.

“I’ve always been into solving puzzles, and with databases, you’re free to explore different parts of the data processing pipeline and apply different skill sets,” he said.

With Arulraj’s arrival in fall 2018 as the Barry Dickman Early Career Professor, he strengthened an already robust data science team in the School of Computer Science. Professor Shamkant Navathe designs and models databases. Assistant Professor Xu Chu focuses on data cleaning. Arulraj complements the others with his emphasis on data processing on modern hardware.

Building non-volatile memory database systems

“Hardware keeps changing all the time, so it’s important to revisit the design decisions you make in software systems,” he said.

Arulraj explored this in his doctoral work at Carnegie Mellon University on non-volatile memory database systems. In the past, databases have relied on two types of storage technologies: fast but volatile memory and slow but durable storage. Recently, however, device manufactures created a new class of memory technologies that are both fast and persistent, blurring the gap between memory and storage. These technologies invalidate many design assumptions that computer scientists have had for decades. Arulraj plans to fill the gap by redesigning database systems to work with these new protocols.

“You were constrained by the characteristics of traditional memory and storage technologies for a long, long time, but finally you have something new,” Arulraj said. “Five years ago, working with these new technologies was like a moonshot, but now it’s a burgeoning field in computer science.”

Next-generation multimedia database systems

Since he arrived at Tech, Arulraj is refocusing his research on video analytics. Multimedia databases are challenging because data is unstructured. With the right tools, though, a user could automatically analyze their photostream to determine the last time they saw a friend or went hiking.

Recent advances in machine learning and graphics processing units have increased the accuracy and speed of video analytics. These developments enable researchers to build systems that can query media, like the one Arulraj is currently working on.

Arulraj’s first course at Tech reflects this research. The seminar combines data analytics and deep learning, though he will regularly teach database system implementation in the spring. The core course teaches how to build data processing systems, covering the whole pipeline of processing data.

“This course will be a much more modern take on database systems research,” he said.

Modern databases cross disciplines, and this was one of the reasons Arulraj chose Georgia Tech.

“There are more opportunities here to collaborate across the board, and that’s the future of database systems.”