UC ECE 945: High Performance Distributed Computing

Instructor:

Philip A. Wilsey
Office: 897 Rhodes   Phone: 556-4779
Office Hrs: Tues/Thurs (all day)
Web site: http://www.ece.uc.edu/~paw/classes/ece945/

Catalog Data:

20-260-945: HIGH PERFORMANCE DISTRIBUTED COMPUTING (HPDC)
This course is devoted to the study of high performance distributed computing (HPDC) and the techniques used in designing and implementing HPDC systems. The course will cover a broad range of issues in distributed computing, including parallel/distributed computing techniques that are applied to the solution of computationally intensive applications across networks of computers.

Prerequisites:

Introductory computer architecture; discrete mathematics; data structures; some familiarity with distributed systems and a working knowledge of at least one high level, block-structured programming language.

Readings:

The basic course will involve a study of the literature on HPDC. The class will be subdivided into five main study areas, namely: (i) introduction to high performance distributed computing, (ii) software performance monitoring techniques, (iii) hardware performance monitoring techniques, (iv) hybrid (software and hardware) performance monitoring techniques, and (v) instrumentation and control/steering of distributed applications.

Although this is a readings class and the topics/papers can be somewhat fluid, we will first cover several basic papers in this topic. In particular, we will review the area of high performance distributed computing. In recent years, high speed networking and improved microprocessor performance are making local and wide networks of workstations (clusters and grids) the ideal choice for cost-effective parallel computing. Our initial focus will be on both hardware and software aspects of computing on clusters and grids. The class will be split into groups with each group responsible for covering and presenting specific topics in the literature. Sample guidelines for making oral presentations are available.

Having familiarized ourselves with the platforms that are typically used for HPDC, we will then look at software/hardware techniques used to improve performance on such distributed execution platforms. We will review several software/hardware performance monitoring techniques and will study in depth the implementation of such techniques. Finally, we will study the applications of software/hardware performance monitoring techniques in distributed applications. Specifically, we will study the use of these techniques in the instrumentation and control/steering of distributed applications.

Of course, this is all subject to your approval. There are many ways that this class can be taken; we could study analytical models of performance of distributed systems, practical considerations for implementing HPDC systems, review the analyses of applications to determine their suitability for distributed computing, or look at other alternatives. A list of papers/books that we might review is included. Please note that we are not confined to this list of papers. You are encouraged to add other papers to this list that you think are of interest to our discussions.

Philip A. Wilsey E-mail: philip.wilsey@uc.edu
Experimental Computing Laboratory Tel: (Off) 513-556-4779
Department of ECE/University of Cincinnati Fax: (Off) 513-556-7326
Last updated