Pinchao Liu

School of Computing and Information Sciences


Lecture Information:
  • April 30, 2019
  • 9:30 AM
  • CASE 349

Speaker Bio

Pinchao Liu is a Ph.D. candidate working on research topics about Cloud Computing and Big Data under the supervision of Dr. Liting Hu. His research interests lie within performance optimization in container cloud and distributed file system. Pinchao received his B.S. and Master’s degrees in Computer Science from TJCU and TUST, China.

Description

In the era of cloud computing, Container as a Service (CaaS) clouds have gained wide-spread prominence. The containerized applications performance and resource utilization could be optimized by using the information of server load, network topology or links. However, it is hard to use these optimization methods into the public clouds shared by multiple tenants, because of the opacity between the tenants and the cloud service providers. Consequently, the containers could be deployed onto the wrong nodes that could result in performance degradation, and to avoid this to happen, resource over-provisioning becomes a general way to deal with it.

My proposed thesis aims at taking a black box approach for finding container ensembles, leveraging deep learning methodology to predict the workload pattern and using overlay network technology to decentralize containers management. For my doctoral dissertation, I propose to develop a framework that uses a combination of solutions such as hierarchical clustering, long short-term memory (LSTM) of the recurrent neural network (RNN) and distributed hash table (DHT) to effectively improve the container cloud resources usage performance.