Communications of the ACM
Tech Helps Companies Detect and Respond to Cloud Computing Performance Bugs
Credit: ENISA
In late 2012, Xiaohui (Helen) Gu presented a research paper on a new tool designed to prevent disruptions in cloud computing. Less than four years later, she has launched a start-up to help companies that use cloud computing improve the user experience for their customers.
"The technology I developed has evolved significantly since 2012," says Gu, an associate professor of computer science at NC State University. "What was then a proof-of-concept is now a technology that works seamlessly with real-world cloud systems, like Amazon Web Services, with a click of a button."
Gu's technology allows companies that use cloud computing to gain insight into user and program behavior in order to diagnose potential problems in their code, so that it can be corrected quickly. The technology is described in "UBL: Unsupervised Behavior Learning for Predicting Performance Anomalies in Virtualized Cloud Systems," presented at ICAC 2012, The 9th ACM International Conference on Autonomic Computing. The technology can prevent performance disruptions in cloud-hosted applications by automatically identifying and responding to potential anomalies before they can develop into disastrous service outages.
The ICAC 2012 paper is authored by Daniel J. Dean, Hiep Nguyen, and Xiaohui Gu.
"What we're doing is important, because services like Spotify and Snapchat are constantly deploying updates to their code," Gu explains. "Our technology can help these companies prevent or detect problems in the updated code and, ultimately, improve the user experience.
"I launched my start-up, InsightFinder Inc., to make this technology available to companies — like Snapchat or Netflix — that use the cloud to deliver their services," Gu says.
In 2013, NC State's Office of Technology Transfer filed for a patent on the technology. Gu started InsightFinder last November.
The start-up was made possible by a U.S. National Science Foundation Small Business Innovation Research grant that Gu is using to develop, test, and fine-tune prototypes.
"Our first public-facing prototype — our beta version — will be available for free, for limited testing, by the end of January," Gu says. "But ultimately, we'll be offering three subscription-based services to corporate customers."
InsightFinder's three services are: a basic cloud-monitoring service; a premium smart monitoring service, which Gu says is more powerful and accurate than existing products on the market based on real-world application testing with a pilot customer; and a batch data analysis service that allows customers to upload historical data to get insights on what caused a performance problem in the past.
"We have a good team in place and are now recruiting customers," Gu says.
No entries found