Using a Unified Approach to Reduce Storage I/O for Big Data Workloads
Previous work in distributed execution engines has focused on reducing query execution time and resource utilization by providing solutions like adaptive data partitioning, data skipping, or caching intermediate data, in isolation. Although these have been majorly studied as different types of solutions, all of them try to exploit the same trade-off: rely on additional/existing storage space to reduce storage I/O and computational overhead.
In this talk, we propose a prototype architecture which unifies all three solutions into one simple framework, and discuss the potential benefits from this architecture.
Theodoros Gkountouvas is currently working for Futurewei Technologies as a Senior Software Engineer at Boston Intelligent Computing and Data Lab. He holds a Bachelor degree in Electrical and Computing Engineering from National Technical University of Athens and M.Sc. degree in Computer Science from Cornell University. His work focuses on designing and implementing solutions to enable and accelerate data analysis.
Mon 17 AugDisplayed time zone: London change
20:00 - 22:00 | Session 2ACSOS In Practice at ACSOS In Practice Meeting Room Chair(s): K R Jayaram IBM Research, USA, Christopher Stewart The Ohio State University, USA | ||
20:00 55mIndustry talk | Challenges in Moving from Datasets to Live Data for Visual Machine Learning ACSOS In Practice | ||
21:00 25mIndustry talk | Leveraging Data Mesh to Optimize Hybrid Cloud for Adaptive AI Control in Industrial Systems ACSOS In Practice Anthony Hill Adapdix Corporation | ||
21:30 25mIndustry talk | Using a Unified Approach to Reduce Storage I/O for Big Data Workloads ACSOS In Practice | ||
