Dec. 9 — Just as Jedi Master Obi-Wan Kenobi could sense disturbances in the Force, computers can sense disturbances in their own operating systems.

But sometimes detecting true disturbances in the “noise” of numerous background tasks run by high-performance computing systems is more difficult because of variance in how tasks are performed. The greater the variation, the greater the inability to run programs efficiently or detect dangerous cybersecurity breeches brought on by malware.

A Virginia Tech researcher has received funding to study variability in high-performance computing, an issue that has broad ramifications.

Variance can affect everything from how efficiently Wall Street executes nearly two billion trades per day, to the maintenance of thousands of e-commerce transactions made at retail giants like Amazon, to securing millions of networks against cyberattacks.

“Variability itself is somewhat of an abstract concept, but in a very real way, variability threatens profit margins by eating into the bottom lines of companies when they have to compensate for potential swings in the performance of datacenters and their interaction with the cloud,” said Kirk Cameron, professor of computer science in Virginia Tech’s College of Engineering and principal investigator. “System variability also impairs our capacity to separate malware from normal system activity because cyberattacks can hide in operating system noise.”

Operating system noise is defined as any interruption of a running program by the operating system. When systems go rogue and are hacked, it’s harder to see the divergence of normal activity when noise is present.

Cameron and his interdisciplinary team are conducting research that will increase the ability of the information technology industry to identify, and also manage, high-performance computer variation peaks.

The team will do this using advanced mathematics and statistical methods to model the impact of variables, such as  power level, time of day, file sizes associated with tasks, and how often memory is accessed.

The collaborative project is funded by the National Science Foundation and draws on several disciplines that include experts in cloud and parallel computing, statistics, and cybersecurity. The primary goal of the project is to develop software, called VarSys, to improve the design and operational efficiencies of both high-performance and cloud systems. The broader impact of VarSys research will apply variability identification and management to improve malware detection.

Daphne Yao, associate professor of computer science, will work on the cybersecurity aspect of the grant.

“How variability impacts security has not been systematically studied,” said Yao. “Noise appears to be a double-edged sword in security. It could make a system more challenging to defend, but it could also make it harder for a hacker to launch certain attacks. We are extremely excited about this project because of the interdisciplinary team and its broad research scope. It gives us a unique opportunity to connect the dots from completely different paradigms,” said Yao.

In addition to Cameron and Yao, the VarSys project team consists of Layne Watson, professor of computer science; Yili Hong, associate professor of statistics in the College of Science; and Ali R. Butt, professor of computer science.

Computer networks, like the Force, will continue to surround and bind us through smart phones, personal computers, and cloud computing devices. And like Jedi masters, Cameron and his team will make sure the disturbances to those networks never let them go dark.

---

Source: Virginia Tech