讲座题目：IBIS: Interposed Big-data I/O Scheduler
Associate professor of the ArizonaState University (ASU) School of Computing, Informatics,and Decision Systems Engineering (CIDSE)
Big-data systemsare increasingly shared by diverse, data-intensive applications from differentdomains. However, existing systems lack the support for I/O management, and theperformance of big-data applications degrades in unpredictable ways when theycontend for I/Os. This talk will introduce IBIS, an Interposed Big-data I/OScheduler, to address this challenge and provide I/O performancedifferentiation for competing applications in a shared big-data system.
IBIS transparentlyintercepts, isolates, and schedules an application’s different phases of I/Osvia an interposition layer on every datanode of the big-data system. Itprovides a new proportional-share I/O scheduler, SFQ(D2), to allow applicationsto share the I/O service of each datanode with good fairness and resourceutilization. Moreover, it enables the distributed I/O schedulers to coordinatewith one another and achieve proportional sharing of the big-data system’stotal I/O service in a scalable manner. The talk will also share theexperimental results of IBIS implemented for Hadoop/YARN, a widely usedbig-data system. The results show that IBIS delivers much stronger performanceisolation than native Hadoop/YARN (99% better for WordCount and 15% better forTPC-H queries) with good resource utilization.
Ming Zhao is anassociate professor of the Arizona State University (ASU) School of Computing, Informatics, andDecision Systems Engineering (CIDSE), where he directs the researchlaboratory for Virtualized Infrastructures, Systems, and Applications(VISA, http://visa.lab.asu.edu). His research is in the areas ofexperimental computer systems, including distributed/cloud, big-data, and high-performancesystems as well as operating systems and storage in general. He is alsointerested in the interdisciplinary studies that bridge computer systemsresearch with other domains. His work has been funded by the NationalScience Foundation (NSF), Department of Homeland Security, Departmentof Defense, Department of Energy, and industry companies, and his researchoutcomes have been adopted by several production systems in industry. Dr.Zhao has received the NSF Faculty Early Career Development (CAREER) award,the Air Force Summer Faculty Fellowship, the VMware Faculty Award, andthe Best Paper Award of the IEEE International Conference onAutonomic Computing. He received his bachelor’s and master’s degreesfrom Tsinghua University, and his PhD from University of Florida.