Invited Keynote Speakers

 

Internet-based Virtual Computing Environment

        Wang Huaimin, Professor

     National University of Defense Technology, China

 

About the Speaker

 

Huaimin Wang received his Ph.D. in computer science from National University of Defense Technology (NUDT) in 1992. He is a Chang Jiang Scholars Program Professor of Ministry of Education, Distinct Young Scholar of NSFC, assembly member of Department of Information Science of NSFC and Fellow of CCF. As the Vice Dean of School of Computer Science and Technology of NUDT, he has worked as the director of several grand research projects, including Internet-based Virtual Computing Environment iVCE)” (funded by 973 Program), “Trustworthy Software Production Tools and Integration Environment” (funded by 863 Program), and  “Research on Trustworthy Software Technology” (funded by NSFC). He has published more than 100 research papers on international conferences and journals of related research fields and has been awarded National Awards for Science and Technology in 2003. His current research interests include distributed system and middleware, Internet computing, trustworthy computing and information security.

Abstract

    Different from the traditional resources, resources over the Internet have such natural characteristics as growth, autonomy and diversity, which have brought grand challenges to their efficient sharing and comprehensive utilization. As there are the essential differences between the Internet computing environment and the traditional computer environment, the resource management methods for traditional computers is no longer suitable for the internet any more. The emergence and development of Web 2.0 and P2P technologies in recent years manifests that the resource management over Internet should adapt to the natural characteristics of the Internet resources and virtualization is one of the potential means to deal with those challenges. However, virtualization in Internet computing environment is essentially different from that in traditional computer environment. There are three scientific issues to be explored, i.e., on-demand aggregation in the open Internet environment, autonomic collaboration among distributed autonomous resources, computational properties of the aggregation and collaboration. We propose the Internet-based Virtual Computing Environment (iVCE), which aims at providing harmonious, trustworthy and transparent computing environment and services for end-users and applications by the means of novel computing models and innovative mechanisms, such as the on-demand resource aggregation and autonomic resource collaboration. The iVCE tries to virtualize and autonomicize Internet resources, design new models and mechanisms adapting well to the Internet computing environment, and promote the sharing and utilization of the Internet resources. Many progresses have been achieved in the model, technical framework and mechanisms of iVCE in the past several years. The concept model of iVCE includes autonomic element (AE), virtual commonwealth (VC) and virtual executor (VE). The architecture of iVCE is designed based on these models. Many mechanisms have been proposed in such areas as resource virtualization, on-demand aggregation, autonomic collaboration, trustworthy enhancement and programming language facility. The iVCE middleware has been designed and implemented, providing a set of services to support the flexible sharing and collaboration of Internet resources. The iVCE middleware has also provided some tools, such as virtual network storage, virtual network memory and virtual cluster for users. Many applications have been run on the iVCE middleware. iVCE will support the multi-scale sharing and collaboration of resources in various granularities, including the large-scale data center, client resources and various resources on the edge of Internet. The productivity and trustworthiness of iVCE will be the focus of investigation in the future. The integration of multiple virtual commonwealths is also part of our future work.


 

The parallel software technology in the new age of computing                          

Dr. Jesse Fang

Managing Director of Intel Labs China

About the Speaker

Jesse Fang is the Managing Director of Intel Labs China, responsible for managing research activities in Intel Labs China, and driving strategic technology collaborations with Chinese industry, academia and government. In 1995, Dr. Jesse Fang joined Intel and founded the Programming System Lab at Intel Labs (formally the Corporate Technology Group). He has been leading the lab to develop programming environment technologies to enable Intel HW uArch research and microprocessor design, and to transfer SW technologies to Intel’s Software Solution Group. Jesse Fang got his Ph.D. in Computer Science at University of Nebraska-Lincoln before he did a post-Doctorate at University of Illinois Urbana-Champaign. He got his B.S. in Math at Fudan University in Shanghai.

Abstract

    With the evolution of the many-core technology offering a higher performance at a lower power consumption, the emerging applications raise new challenges to the software technology. Parallel computing is not only the backbone of data centers and super computers, it also moves into the core of embedded systems that support daily life applications. Based on the decades of research experience in microprocessors, now researchers from Intel labs are exploring the new territory of software technology. Focused on embedded system research, in addition to the research on embedded software, Intel labs China also collaborated with local government, academia and industry to support the China's leadership in parallel software development by the Intel China Center of Parallel Computing (ICCPC).


 

Building a Domain-Knowledge Guided System Software Environment to Achieve High-Performance of Multi-core Processors

Xiaodong Zhang, Robert M. Critchfield Professor in Engineering

The Ohio State University, USA

About the Speaker

Xiaodong Zhang is the Robert M. Critchfield Professor in Engineering, and Chairman of the Department of Computer Science and Engineering at the Ohio State University. His research interests cover a wide spectrum in the areas of high performance and distributed systems. Several technical innovations and research results from his team have been widely adopted in commercial processors, major operating systems and databases, making direct contributions to the advancement of memory systems.He received his Ph.D. in Computer Science from University of Colorado at Boulder, and his B.S. in Electrical Engineering from Beijing University of Technology. He is a Fellow of the IEEE.

 Abstract

       Although multi-core processors have become dominant computing units in basic system platforms from laptops to supercomputers, software development for effectively running various multi-threaded applications on multi-cores has not made much progress, and effective solutions are still limited to high performance applications relying on exiting parallel computing technology. In practice, majority multi-threaded applications are highly concurrent programs demanding high throughput, such as concurrent database transactions, massive and independent query requests in Web servers and search engines, and executing many-tasks for scientific applications in a multiprogramming mode. To best utilize the increasingly rich computing and cache resources in multi-core processors (many-cores in the near future) we must address several serious and difficult challenges. First, there are several critical hardware resources for multi-threads to share, such as the last level caches and the memory buses. However, the shared resource management is Largely controlled by hardware. Second, OS scheduler has little knowledge about applications' data demanding and access behavior, making sub-optimal task assignment decisions. Finally, the space allocation for each thread in the shared-cache is demand-based, often causing access conflicts and pollution, significantly degrading overall execution performance.

We have developed a runtime environment connecting multi-threaded applications and system software in a collaborative way, where operating system is guided by application domain knowledge including data access locality and execution behavior to schedule tasks and allocate shared hardware resources for each running thread. We evaluate our environment by concurrent database transactions and multi-threaded scientific computing programs, and show strong performance and Throughput improvement by minimizing cache conflicts and misses in the last level caches of multi-cores. We further develop our system as a general framework to automatically manage multi-threaded applications on multi-core processors.


We strongly recommend you to view the site with Mozilla Firefox.


For any question please don't hesitate to contact zyshao.hust@gmail.com (email subject "NPC2010 Question").

Copyright NPC 2010. Created and Maintained by NPC2010 Web Team.
Mirror URL: http://grid.hust.edu.cn/npc2010