Title: Runtime Optimizations for Large-Scale Data Analytics
Advisor: Magdalena Balazinska
Supervisory Committee: Magdalena Balazinska (Chair), Bingni Brunton (GSR, Biology), Dan Suciu, Alvin Cheung, and Arvind Krishnamurthy
Abstract: Large-scale data analytics benefits from optimizations from many aspects. One category of optimizations is hard to be performed statically since they rely on information that is only available during runtime. We show that these runtime optimizations have large impact on application performance. Specifically, we investigate the problem from the following aspects: execution models for iterative queries in shared-nothing architectures, elastic memory management for cloud data analytics, and real-time visual applications. We summarize our work on the first two subjects and propose our plan for the third project.