VarSys is a project sponsored by the U.S. National Science Foundation (Grant NSF CISE CSR #1565314 beginning August 1, 2016). The PI is Kirk W. Cameron of Virginia Tech. Virginia Tech Co-PI’s include: Ali Butt (Cloud Systems), Daphne Yao (Security), Yili Hong (Statistics), Layne Watson (Mathematics/Algorithms), and Godmar Back (Computer Systems Education). Other participants include Margaret Ellis (Outreach and Education) and Amy Loeffler (Communications Manager) and more than a dozen graduate and undergraduate students from several sub areas and disciplines.
In this project, we are building a software framework (VarSys) to identify, characterize, isolate, and manage variability in high-performance parallel and distributed systems. Our overarching goal is to create fundamental technologies that map system and application configurations to variability. The resulting variability maps will provide researchers a scientifically rigorous framework to avoid relying solely on experience or intuition to manage variability. We are using the VarSys framework to improve HPC experimental design; improve cloud services; and improve malware detection. We are conducting community outreach activities that introduce variability into undergraduate operating system courses leveraging a variability job submission system and a variability visualization tool. A team of undergraduates are creating new protocols for classifying systems for variability — these protocols will be used in upcoming public outreach activities.