Faculty Candidate #2
- To: Computer Science
- Subject: Faculty Candidate #2
- From: Krishnan Pillaipakkamnatt <Krishnan.Pillaipakkamnatt@Hofstra.edu>
- Date: Wed, 28 Mar 2018 06:26:36 -0400
- Cc: CSC Faculty, CSC Faculty PT, Krishnan Pillaipakkamnatt <Krishnan.Pillaipakkamnatt@Hofstra.edu>
Our second faculty interview will be held today. I urge you to attend the presentation, and also take the opportunity to meet with the faculty candidate after the presentation. The department welcomes your input in this process.
“Jianchen Shan is currently a Ph.D. Candidate in the Department of Computer Science at New Jersey Institute of Technology. His research interests and experience span the areas of Cloud Computing, Parallel and Distributed Computing, High-Performance Computing, Mobile Computing, Big Data, Operating System and Computer Architecture. His work on Mobile Distributed Cloud Computing is widely covered in the press, such as ABC News, Fox News, NJ Business Magazine and The Times, etc.”
Adams Hall 208
Pizza will be served
High Performance Cloud Computing on Multicore Computers
Abstract: Cloud has become a major computing platform, in which applications run in virtual machines (VMs) hosted on multicore servers. Most VMs in the cloud have multiple virtual CPUs running multiple threads to leverage the computing power of multiple cores. The number of VCPUs in a VM keep increasing, in accordance with the growing core count in a physical server.
The research focuses on the performance of multi-threaded executions in the cloud, identifies the causes of performance issues, and develops innovative techniques to improve performance. A multi-threaded execution consists of multiple threads collaborating together to finish computation tasks. Most of the existing efforts target the performance of individual threads and have significantly reduced the overhead of single thread execution. However, how virtualization and resource sharing affect the overall performance of a set of the collaborative thread is understudied.
The research first concentrates on the interaction between threads. It experimentally examines the overhead caused by virtualization on the interaction between threads and the performance indications. The research also develops a technique named APPLES to substantially reduce the overhead caused by spinning-based coordination in VMs. Then, the research studies the performance impact of CPU resource sharing on the performance of multi-threaded executions. To improve the overall performance of multi-threaded execution with shared CPU resource, it proposes to dynamically adjust the distribution of CPU resource to different VCPUs based on the resource demand of the threads on the VCPUs.
Department of Computer Science