Dr. Ming Zhao’s research interests are in experimental computer systems, especially systems for virtualized/cloud computing, high-performance computing, and autonomic computing.
Virtualization is an enabling technology for creating important new system abstractions and addressing the various challenges faced by today’s computing systems. The use of virtualization can span across computing systems of different sizes, from desktops to supercomputers, and across different layers of the systems, from virtual machines, virtual storage, to virtual networks. The fundamental goal of Dr. Ming Zhao and his VISA Research Lab is to explore innovative techniques in virtualization as well as autonomics in order to effectively utilize the resources in large-scale, dynamic, and complex computing systems and to support the high-performance, robust, and secure computing of challenging applications from different domains. Currently, his team is focusing on several projects related to virtualization in distributed and high-performance computing systems.
System virtualization is a powerful technology that enables the emerging computing paradigms such as public and private cloud systems. It allows applications to be conveniently deployed along with their required execution environments through virtual machines (VMs), and supports them to flexibly share the underlying physical resources with strong isolation. However, there exists an increasingly urgent need for virtualized systems to deliver strong Quality of Service (QoS) guarantees to their hosted applications. Currently such systems can meet only coarse-grained and relaxed performance requirements, and their management considers only limited facets of an application’s multi-type resource usage. The continued existence of the lack of strong QoS guarantees from virtualized systems presents a critical hurdle to their further adoption by applications with diverse QoS requirements and their support of more economical QoS-based charging models. The objective of this FIU and DHS sponsored project is to create a QoS-driven multi-type resource management system and to support strong QoS guarantees for applications hosted on virtualized computing systems. Dr. Zhao and his team are creating novel nonlinear modeling and predictive control based techniques for accurate and adaptive on-demand resource allocation as well as new cross-layer optimization techniques for improved application QoS and resource-use efficiency (Fig. 1). The proposed research is significant because it will enable virtualized systems to support strong QoS guarantees and efficient resource allocation for applications with dynamic and complex multi-type resource usage behaviors. As a result, this project will allow a broader range of applications with different levels of QoS requirements to benefit from the use of virtualized computing. It will enable virtualized systems such as public clouds to provide QoS-based contracting model, which will allow resource providers to more efficiently allocate their resources across VMs and allow users to more cost-effectively purchase resources for their applications.
High-performance computing (HPC) systems are important platforms for solving challenging problems in many science and engineering domains. In such systems, the parallel file system (PFS) is at the core of the storage infrastructure, which provides applications high-throughput data access through the parallelism of I/Os. However, existing PFS-based storage systems are unable to recognize the different application I/O workloads and incapable of satisfying the applications’ different I/O bandwidth needs. These limitations prevent applications from efficiently utilizing the HPC resources while achieving their desired QoS. This problem is continuing to grow with the ever-increasing scale of HPC systems and with the increasing complexity and number of applications running concurrently on these systems. It presents a hurdle for the further scale-up of HPC systems to support many large, data-intensive applications. The objective of this NSF-funded project is to address the research challenges in application QoS-driven storage management in HPC systems, in order to support the allocation of storage resources on a per-application basis as well as the efficient execution of applications with I/O demands varying by several orders of magnitude. Dr. Zhao and his team are creating new parallel file system virtualization and autonomic bandwidth allocation techniques to enable per-application QoS-driven control and optimization of HPC storage resources (Fig. 2). This research has the potential to drastically improve the state of the art in I/O management in existing HPC systems and generate an impact on the design of future systems. It will also allow HPC applications with diverse I/O characteristics and requirements to achieve their desired QoS on shared HPC resources.
- National Science Foundation, Department of Homeland Security, and industry
- Excellence in Student Mentoring, Florida International University School of Computing and Information Science, 2012
- Best Student Paper Award of the 7th International Conference on Autonomic Computing (ICAC), 2007.
- Conference organizations:
- Program Chair: The International Workshop on Feedback Computing
- Technical Program Committee: The IEEE International Conference on Big Data, The IFIP International Conference on Network and Parallel Computing, The ACM International Symposium on High-Performance Parallel and Distributed Computing, The International Workshop on Virtualization Technologies in Distributed Computing, The International Conference on Autonomic Computing, The IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing, The International Conference on Computer and Management, Workshop on Management of Cloud Systems, The IEEE International Conference on Parallel and Distributed Systems, The International ICST Conference on Sensor Systems and Software, The IEEE/IPSJ International Symposium on Applications and the Internet, The International Workshop on System Management Techniques, Processes, and Services, Workshop on Mirco Architectural Support for Virtualization, Data Center Computing, and Clouds, The International Conference on Autonomic and Trusted Computing
- Organizing Committee: The International Conference on Autonomic Computing, The International Conference on Software Engineering and Knowledge Engineering, The International Green Computing Conference, The IEEE International Workshop on Storage Network Architecture and Parallel I/Os, The International Conference on Autonomic and Trusted Computing, The International ICST Conference on Sensor Systems and Software, The Annual Meeting of the IEEE/ACM International Symposium on Modeling, Analysis and Simulation of Computer and Telecommunication Systems, The International Workshop on Virtualization Performance: Analysis, Characterization and Tools, The International ICST Conference on Cloud Computing
- IBM Research, HP Labs, Marvell Semiconductor, CloudVPS, Fusion-io, Sandia National Labs, Los Alamos National Labs
- S. Liu, S. Ren, G. Quan, M. Zhao, and S. Ren, “Profit Aware Load Balancing for Distributed Cloud Data Centers”, Proceedings of the 27th IEEE International Parallel & Distributed Processing Symposium (IPDPS2013), May 2013. (Acceptance Rate: 21%)
- R. Koller, L. Marmol, R. Rangaswami, S. Sundararaman, N. Talagala, M. Zhao, “Write Policies for Host-side Flash Caches,” Proceedings of the 11th USENIX Conference on File and Storage Technologies (FAST’13), February 2013. (Acceptance Rate: 18%)
- D. Arteaga, M. Zhao, P. V. Riezen, and L. Zwart, ”Trace Analysis for Block-level Caching in Cloud Computing Systems,” Work-in-progress paper of the 11th USENIX Conference on File and Storage Technologies (FAST’13), February 2013.
- Y. Xu, A. Suarez, and M. Zhao, “IBIS: Interposed Big-data I/O Scheduler,” Work-in-progress paper of the 11th USENIX Conference on File and Storage Technologies (FAST’13), February 2013.
- L. Wang, J. Xu, M. Zhao, “Modeling VM Performance Interference with Fuzzy MIMO Model,” Proceedings of the 7th International Workshop on Feedback Computing (FeedbackComputing, co-held with ICAC2012), September 2012.
- L. Wang, J. Xu, M. Zhao, “Application-aware Cross-layer Virtual Machine Resource Management,” Proceedings of the 9th International Conference on Autonomic Computing (ICAC2012), September 2012. (Acceptance Rate: 24%)
- Y. Xu, D. Arteaga, M. Zhao, Y. Liu, R. Figueiredo, S. Seelam, “vPFS: Virtualization-based Bandwidth Management for Parallel Storage Systems,” Proceedings of the 28th IEEE Conference on Massive Data Storage (MSST’12), April 2012. (Acceptance Rate: 24%)