A Survey of Scheduling in Cluster Computing

PhD Qualifying Examination


Title: "A Survey of Scheduling in Cluster Computing"

by

Mr. Chen CHEN


Abstract:

With the prevalence of large scale data analytics, it has become a norm to run 
the data parallel applications in a large cluster of machines. Having various 
applications coexisting in a cluster, the cluster scheduler serves as a 
critical component to the overall system performance and service quality. An 
idealized scheduler shall be general enough to accommodate multiple cluster 
computing frameworks, like Hadoop and Spark. Meanwhile, the scheduler in 
production clusters shall scale well to provide timely response when massive 
scheduling requests keep arriving. More importantly, predictable and fast job 
response is highly expected to ensure good user experience of end-user-facing 
products like Google search, and this requires cluster schedulers to provide 
both fairness and performance. As a result, recent years have witnessed 
unremitting research efforts for designing appropriate cluster schedulers 
satisfying the above requirements.

This survey provides a systematical review about state-of-the-art cluster 
schedulers. Besides generality and scalability, we mainly focus on the fairness 
and performance aspects. Albeit conflicting, fairness and performance are both 
desirable features for a cluster scheduler, and striking a balance between them 
is practically necessary. Nonetheless, current solutions for that compromise 
are merely complex heuristics, without theoretical support or worst-case 
guarantees. Further explorations are called for on that problem.


Date:			Thursday, 9 March 2017

Time:                  	2:00pm - 4:00pm

Venue:                  Room 3494
                         Lifts 25/26

Committee Members:	Prof. Bo Li (Supervisor)
 			Dr. Wei Wang (Supervisor)
 			Prof. Lei Chen (Chairperson)
 			Dr. Kai Chen


**** ALL are Welcome ****