Path: uflorida!simulation From: simulation@uflorida.cis.ufl.edu (Moderator: Paul Fishwick) Subject: SIMULATION DIGEST V32 N2 Newsgroups: comp.simulation Reply-To: simulation@uflorida.cis.ufl.edu Sender: fishwick@uflorida.cis.ufl.edu Approved: fishwick@uflorida.cis.ufl.edu Distribution: world Volume: 32, Issue: 2, Sun Mar 21 21:24:31 EST 1993 +----------------+ | TODAY'S TOPICS | +----------------+ [NEW QUESTIONS] Systems Dynamics Implementations Time Warp Code Availability of NEST H-Fields Block Model Simulator [CALL FOR PAPERS/PARTICIPATION] Distributed Collaboration Virtual Reality Short Course Modeling of Computer Systems * Moderator: Paul Fishwick, Univ. of Florida * Send topical mail to: simulation@bikini.cis.ufl.edu OR post to comp.simulation via USENET * Archives available via FTP to bikini.cis.ufl.edu (128.227.224.1). Login as 'anonymous', use your e-mail address as the password, change directory to pub/simdigest. Do 'binary' before any file transfers. * Simulation Tools available by doing above and changing the directory to pub/simdigest/tools. * Gopher link, containing digest and software archives, is available at gopher.cis.ufl.edu. ----------------------------------------------------------------------------- Subject: [NEW QUESTIONS] ------------------------------ Date: 17 Mar 93 08:56:00 EST From: "DRCOA1::JLUNA" Subject: Systems Dynamics Tools and Psychology Simulations Subject: Systems Dynamics Tools and Psychology Simulations Hello! I'm looking for current implementations of Systems Dynamics modeling and simulation that can run in a 386/486 DOS or Windows environment. I have a couple of outdated tools that run on DOS, but what I'm interested in doing is getting a hold of an existing environment and augmenting it as needed (so a Windows product is preferred I guess). In any case, an existing tool would be useful. I'm familiar with Stella on the Mac, but I'm having trouble locating High Performance Systems. Also (out of curiosity), is DYNAMO still around? (Francois, are you out there?) Also, does anyone know of 'macro' simulation models for psychology? That is, not models of brain function (neural nets, cellular automata, etc.) per se, but more along the lines of a systems approach (with feedback, possibly) to psychological processes (including but not limited to learning), or perhaps using chaos or some kind of stochastic rule-based approach. I am already familiar with Lewin's equational approach if that helps at all. I realize this is vague, but any and all leads will be appreciated. Please respond directly regarding the first request if it is convenient. Thanks! Joel Luna Dynamics Research Corporation 60 Frontage Rd Andover, MA 01810 (508) 475-9090, x1368 Fax (508) 470-0201 e-mail: jluna%drcoa1.decnet@drcvax.af.mil ------------------------------ Date: Thu, 18 Mar 93 08:54:13 EST From: subhas@cs.wm.edu (Subhas Roy) Subject: Need Time Warp code I need source code for Time Warp (Mother of all PDES :-). Looking for the source code of public domain (free/semi-free) implementation of Time Warp (which is a mechanism for optimistic execution of PDES (parallel discrete event simulation)). I want to use that for my disseration on PDES. Does anybody know the availability of any such code? I will greatly appreciate if somebody provides the source code for Time Warp kernel or pointer to the same. Info about any Time Warp simulator is also welcome. Thanks in advance. Please reply to: subhas@cs.wm.edu Subhas Roy Dept. of Computer Science College of William & Mary, Virginia E-Mail : subhas@cs.wm.edu ------------------------------ From: bhati@plains.NoDak.edu (Amit Bhati) Subject: NESTPACK or other progs. to simu. distributed comp. sys. Help! Date: Mon, 22 Mar 1993 00:53:25 GMT I am looking for information about the NESTPACK simulation package, that was develpoed at Columbia Univ. to simulate networks and distributed computing systems. The article that I gathered from a gopher server seems to be outdated, as all the e-mail contact addresses are ambiguous. The ftp server at host "columbia.edu" does not contain a ~ftp/pub directory in which the program files are supposed to be present. Can anyone help? Or do you know of better software that a penny-less student can use for the same purpose? e-mailed replies are prefered. I am appending the article, just in case it helps or is of interest to you. Cheers, amit bhati bhati@plains.nodak.edu ----------------------------THE ARTICLE------------------------------------- Nest - A Network Simulation Testbed ----------------------------------- Jed Schwartz 450 Computer Science Building Columbia University New York, NY 10027 E-mail:jed@douglass.columbia.edu Phone: (212) 854-8192 Description ----------- The NEST simulation package developed at Columbia provides a generic framework for simulating networks and prototyping distributed systems. The NEST simulation runs in a sin- gle UNIX process, and the NEST monitor runs in a separate process, which may be remote. Simulations are created using the NEST library, a C-language library which runs under all UNIX variants on VAX and 680x0 processors (SPARC and RT soon). The library provides a simple, high- level message-passing facility among a set of concurrent lightweight processes which simulate a distributed system. Simulations can be interactively monitored and managed using the standard Nest monitor program, which permits dynamic reconfiguration of the simulation. It is possible to create/halt/continue/destroy simulated nodes and links, and to alter their function and parameters, while a simulation is in progress. Multiple monitors at different sites can monitor a single simulation. The monitor is a monochrome SUNView application (an X Windows implementation is under development). Network Address: Both documentation and full sources are available on the Columbia Vax [columbia.edu: 10.3.0.89]. There are four files in the ~ftp/pub directory: The information in this section is provided in accordance with the copyright notice appearing at the front of this guide. June 28, 1989 NNSC Section M.5, Page 1 221184 bytes nest-25-doc.tar Lineprinter formatted documentation. 438272 bytes nest-25-doc2.tar Scribe and PostScript formatted docs. 339968 bytes nest-25-src.tar Source for Nest simulation library. 208896 bytes nest-25-disp.tar Source for user interface display. Who Can Use the Service ----------------------- Nest is available without cost to non-profit research and educational organizations which have access to the Inter- net. Others may order a tape through mail for a cost of $100. by contacting Jed Schwartz at the above address. Nest is made available for research purposes only. No part of the Nest software may be redistributed by itself or as part of any software mpackage, except by express arrangement. For-profit organizations and those without Internet access may order a Nest tape through mail for a cost of $100. by contacting Jed Schwartz at the above address. Miscellaneous Information ------------------------- Please direct technical questions to Alex Dupuy: alex@columbia.edu, (212)854-4290 June 28, 1989 NNSC Section M.5, Page 2 -------------------------------------------------------------------------- -- ---------------------------------------------------------------------------- |Amit Bhati bhati@plains.NoDak.edu | "All that is there in the| |Computer Science Dept., !uunet!plains!bhati | middle of the road is a | |IACC Building, NDSU, Ph: (701) 232-6754 | dumb, yellow line and a | |Fargo, ND 58105. | dead skunk." | ------------------------------ From: Joe Desbonnet Subject: H-fields (where?) To: simulation@bikini.cis.ufl.edu Date: Mon, 22 Mar 93 1:50:09 GMT I am looking for a semiconductor simulation package called 'H-fields'. I believe its freely available, but I have not been able to find a FTP site or any other information. Any pointers would be greatly appreciated. Joe. -- Joe Desbonnet, Email: joe@epona.physics.ucg.ie Dept of Physics, phydesbonnet@bodkin.ucg.ie University College Phone: +353-91-24411 x 2716 Galway, Ireland. Fax : +353-91-25700 ------------------------------ Date: Sat, 20 Mar 93 17:32:01 -0500 From: mxb11@po.CWRU.Edu (Marc Buchner) To: simulation@bikini.cis.ufl.edu Subject: Continuous-time Simulation Package I'm looking for a block oriented continuous-time simulation package that is available in the public domain. I am familiar with a primitive version of LEANS (Lehigh Analog Simulator??) that was available many moons ago. [ED: There are C and C++ versions of a block simulator in SimPack and SimPack++ (see the head of this digest under software tools). It is very "bare bones", but the hooks are there to easily add your own transfer functions -PAF] A C++ version of such a simulation package would be ideal. Any help would be appreciated... Marc -- Marc Buchner 617 Crawford Hall Systems Engineering Department Case Western Reserve University Cleveland, OH 44122 phone: (216) 368-4096 fax: (216) 368-3123 ------------------------------ Subject: [CALL FOR PAPERS/PARTICIPATION] ------------------------------ Date: 17 Mar 1993 03:14:45 U From: "Stephen DownesMartin" Subject: Call For Papers To: "Paul Fishwick" Content-Length: 2133 X-Lines: 52 Call For Papers Session: Interactive Simulation Support for Distributed Collaboration Track: Modeling Methodology Conference: Winter Simulation Conference, Los Angeles, December 12 - 15, 1993 This session will examine the issues surrounding the use of distributed interactive simulation as a mechanism for mediating distributed collaboration. Domains of interest include but are not restricted to: Education, Training, Corporate Decision Making, Medical Consultation, Intelligence Applications, Disaster Relief Planning. Issues of interest focus on the modeling methodology (design and execution) of using distributed intelligent software agents as the interface between users and a distributed collaboration system. Design criteria for software agents include their ability to act as personal assistants in the presence of users, replacements in the absence of users, and generalization across a wide range of domains. Execution issues include their ability to utilize heterogeneous tools from a wide range of human domains, distribution over networks of varying bandwidth, and the ability to utilize current and future advances in artificial intelligence techniques. Anyone wishing to present a paper at this session is invited to submit a two page abstract, preferably by e-mail, before April 5, 1993 to: Stephen Downes-Martin David Sarnoff Research Center CN 5300 Princeton, NJ 08543-5300 Tel: 609-497-2620 e-mail: stephen_downesmartin@maca.sanoff.com Thank you, Stephen Downes-Martin --------------------- ------------------------------ Date: Thu, 18 Mar 93 16:10:57 GMT From: burdea@telephone.rutgers.edu (Greg ) To: isslist@ivy.princeton.edu Subject: Virtual Reality Short Course VIRTUAL REALITY Systems and Applications IEEE-ELECTRO'93 International Conference The One-day Technical Short Course will be offered at IEEE-Electro'93. This course will first present key technologies used in Virtual Reality, including head-mounted displays, sensing gloves , 3-D directional sound boards and force/touch feedback tools. Then we will present unified systems that integrate these technologies with state-of-the-art graphics workstations. The second part of the course will present emerging applications of Virtual Reality in the areas of electro-mechanical product design, military simulations, medical diagnosis and training, multi-media teleconferencing, architectural planning and entertainment. Each student will be given a proceedings book and videotape. Lunch will be provided. Instructor: Dr. Grigore Burdea, Rutgers University Date: April 28, 9:30 am-5:30 pm. Place: Ramada Inn, Edison, NJ. Fee: $300 before April 2, $350 after this date. For more information and Registration call 1-800-877-2668 ext. 244 ------------------------------ Sender: owicki@src.dec.com (Susan Owicki) Newsgroups: comp.simulation From: owicki@src.dec.com Subject: SIGMETRICS 93 announcement To: comp.simulation.usenet@inet-gw-2.pa.dec.com Date: Thu, 18 Mar 93 14:45:30 -0800 Title: 1993 ACM SIGMETRICS Conference on Measurement and Modeling of Computer Systems Date: May 10 - 14, 1993 Place: Santa Clara Marriot, Santa Clara, California ------------------------------------------------------------ OVERVIEW The 1993 Annual Conference on Measurement and Modeling of Computer Systems will be held May 10 - 14, 1993, in Santa Clara, California. It will feature - Tutorials (May 10, 11) - Technical Conference with: Keynote Spearkers Presentations of papers Poster and work-in-progress sessions. This on-line announcement contains the following sections: - TUTORIAL SCHEDULE - CONFERENCE PROGRAM - TUTORIAL ABSTRACTS - PEOPLE - LOCATION - CONFERENCE AND TUTORIAL PRE-REGISTRATION - HOTEL RESERVATION ----------------------------------------------------------- TUTORIAL SCHEDULE (see below for abstracts) Extended Tutorials (May 10, 1993) Both Extended Tutorials are presented concurrently from 1:00 - 5:00 Jim Gray - A Tour of Popular DB and TP Benchmarks Fouad Tobagi - B-ISDN: SONET, ATM, and Fast Packet Switching Regular Tutorials (May 11, 1993) Within each track, the times for the 1st, 2nd, 3rd, and 4th tutorials are: 8.30-10:00, 10:30-12:00, 1:30-3:00, and 3:30-5:00. Systems Architecture Alan Smith - Cache Memories. David Kaeli - Issues in Trace-Driven Simulation. Garth Gibson - Performance and Reliability in Disk Arrays. P. Venkat Rangan - Architectures and Algorithms for Digital Multimedia On-Demand Servers. Performance Methodology Kang Shin - Evaluation of Real-Time Fault-Tolerant Computing Systems. Daniel Reed - Performance Environments for Scalable Parallel Systems. C. M. Woodside - Performance Engineering of Client-Server Software Systems. Philip Yu - Modeling and Analysis of Transaction Processing Systems. Abstract Models Richard Muntz - The Stochastic Complement, Decomposition and Aggregation in the Numerical Solution of Markov Models. Guy Latouche - Divide-and-Conquer Methods for Quasi-Birth-Death Processes. Benjamin Melamed - TES Modeling of Autocorrelated Time Series with Applications to Source Modeling of Compressed Video. Joe Hellerstein - Modeling Dynamic Behavior with Time Series Analysis. ------------------------------------------------------------ CONFERENCE PROGRAM ************ Monday, May 10 *********** 12:30 - 1:30 Registration 1:30 - 5:00 Extended Tutorials *********** Tuesday, May 11 ********** 7:30 - 8:30 Registration 8:30 - 5:00 Tutorials 7:00 - 9:00 Reception and registration ************ Wednesday, May 12 ************ 8:45 - 9:00 Opening Remarks 9:00 - 10:00 Keynote Speaker "Software Technologies for Massively Parallel Processing" Lennart Johnsson Thinking Machines Corp. and Harvard University 10:00 - 10:30 Break 10:30 - 11:30 I/O System Performance I Chair: Garth Gibson, CMU A New Approach to I/O Performance Evaluation - Self-Scaling I/O Benchmarks, Predicted I/O Performance Peter M. Chen and David Patterson University of California at Berkeley Trace Driven Analysis of Write Caching Policies for Disks Prabuddha Biswas and K.K. Ramakrishnan, DEC and Don Towsley, University of Massachusetts at Amherst 11:30 - 12:30 Cache Simulation Chair: Anoop Gupta, Stanford University Efficient Simulation of Caches under Optimal Replacement with Applications to Miss Characterization Rabin A.Sugumar and Santosh G. Abraham University of Michigan at Ann Arbor Cache Inclusion and Processor Sampling in Multiprocessor Simulations Jacqueline Chame and Michel Dubois University of Southern California 12:30 - 1:30 Lunch 1:30 - 2:30 Posters 2:30 - 4:00 Parallel Systems I Chair: John Zahorjan, University of Washington The Wisconsin Wind Tunnel: Virtual Prototyping of Parallel Computers Steven K. Reinhardt, Mark D. Hill, James R. Larus, Alvin R. Lebeck, James C. Lewis and David A. Wood University of Wisconsin at Madison The Influence of Random Delays on Parallel Execution Times Vikram Adve and Mary K. Vernon University of Wisconsin at Madison The KSR1: Experimentation and Modeling of Poststore E. Rosti, E. Smirni, T.D. Wagner, A.W. Apon and L.W. Dowdy Vanderbilt University 4:00 - 4:30 Break 4:30 - 5:30 Work in Progress 8:30 Town Meeting ************ Thursday, May 13 ************ 8:30 - 9:30 I/O Systems II Chair: Don Towsley, University of Massachusetts at Amherst The Process-Flow Model: Examining I/O Performance from the System's Point of View Gregory R. Ganger and Yale R. Patt University of Michigan at Ann Arbor An Analytic Performance Model of Disk Arrays Edward K. Lee and Randy H. Katz University of California at Berkeley 9:30 - 10:30 Keynote Address: "Terabytes >> Teraflops (Or why work on processors when I/O is where the action is?)" David Patterson University of California at Berkeley 10:30 - 11:00 Break 11:00 - 12:00 Tools and Techniques Chair: Peter Danzig, USC MEASURE+ - A Measurement-Based Dependability Analysis Package Dong Tang and Ravishankar K. Iyer University of Illinois at Urbana-Champaign On the Sensitivity of Transient Solutions of Markov Models A.V. Ramesh and Kishor Trivedi Duke University 12:00 - 1:00 Lunch 1:00 - 2:00 Posters 2:00 - 3:00 Simulation Methods Chair: Ben Melamed, NEC USA Parallel Simulation of Markovian Queueing Networks Using Adaptive Uniformization David Nicol, College of William and Mary Philip Heidelberger, IBM The Accuracy of Trace-Driven Simulations of Multiprocessors Stephen R. Goldschmidt and John L. Hennessy Stanford University 3:00 - 3:30 Break 3:30 - 5:00 Parallel Systems II Chair: Larry Dowdy Vanderbilt University Processor Scheduling on Multiprogrammed, Distributed Memory Parallel Computers Sanjeev K. Setia, University of Maryland Mark S. Squillante, IBM Satish K. Tripathi, University of Maryland Performance Comparison of Thrashing Control Policies for Concurrent Mergesorts with Parallel Prefetching Kun-Lung Wu, Philip Yu and James Teng, IBM A Markov-Modulated Bernoulli Process Approximation for the Analysis of Banyon Networks Dikran S. Meliksetian, South Dakota School of Mines and Technology C.Y. Roger Chen, Syracuse University Evening Banquet, Decathlon Club ************ Friday, May 14 ************ 8:30 - 10:00 Real-Time Scheduling Chair: Ian Akyildiz, Georgia Tech Modeling and Validation of the Real-Time Mach Scheduler Hiroshi Arakawa, Daniel I. Katcher, Jay K. Strosnider and Hideyuki Tokuda, CMU ROBUST: A Hardware Solution to Real-Time Overload Sanjoy Baruah, University of Texas at Austin Jayant R. Haritsa, University of Maryland Efficient On-line Processor Scheduling for a Class of IRIS (Increasing Reward with Increasing Service) Real-Time Tasks Jayanta K. Dey, James F. Kurose, Don Towsley, C.M. Krishna and Mahesh Girkar, University of Massachusetts at Amherst 10:00 - 10:30 Break 10:30 - 12:00 Cache Models Chair: Derek Eager, University of Saskatchewan Analysis of Superposition of Streams into a Cache Buffer R.J.T. Morris, IBM Analyzing Multiprocessor Cache Behavior Through Data Reference Modeling Jory Tsai, DEC Anant Agarwal, MIT Effectiveness of Trace Sampling for Performance Debugging Tools Margaret Martonosi and Anoop Gupta, Stanford Thomas Anderson, University of California at Berkeley 12:00 - 12:15 Closing Remarks ---------------------------------------------------------- TUTORIAL ABSTRACTS Monday, May 10, 1:00 - 5:00 A Tour of Popular DB and TP Benchmarks Jim Gray Digital Equipment Corporation 455 Market St. 7th Floor San Francisco, CA 94105 Abstract This class will tour the design, rational, and experience with the popular (and some not so popular) benchmarks used in the database (DB) and transaction processing (TP) communities. Premier among these are the Transaction Processing Performance Council benchmarks (TPC-A and B) and the newly approved TPC-C (the order entry benchmark). These benchmarks measure OLTP performance. The benchmarks and their model of throughput (performance) and price performance will be explained. Experience with the benchmarks explains the rationale for the newer ones. The second lecture discusses benchmarks covering decision support (Wisconsin, Set-Query, TPC-D), OO databases (OO1 and OO7), text retrieval (Full Text Retrieval), and scientific databases (Sequoia). Implicit in the course will be the concepts and techniques to go into benchmark design: simplicity, portability, scaleability, and relevance. Much of the class is based on the new edition of "The Performance Handbook for Database and Transaction Processing Systems". ............................................................. Monday, May 10, 1:00 - 5:00 B-ISDN: SONET, ATM, and Fast Packet Switching Fouad A. Tobagi Department of Electrical Engineering Stanford University Stanford, California 94305 and Starlight Networks Mountain View, California 94041 Abstract Two characteristics mark the design of future telecommunications networks: (i) the ability to support high bandwidth applications (such as digitized video and image communication), and (ii) the ability to integrate all services on the same network. In response to these needs, the switching technique found to be the most appropriate is a form of packet switching, commonly referred to as Asynchronous Transfer Mode (ATM). In this tutorial, we begin with a brief overview of the projected applications and their bandwidth requirements. We then describe SONET, the physical layer underlying ATM networks, and the ATM standard protocols as defined by CCITT and the ATM Forum. Finally, we address the issue of fast packet switching and survey the various architectures used in building switches for ATM networks. The course outline includes, 1. Broadband Applications and Bandwidth Requirements, 2. Fiber Optics Transmission Facilities: SONET, Frame Formats, Data Rates, 3. ATM Standards: Architecture and Protocols. and 4. Fast Packet Switch Architectures, Performance and Implementation. ............................................................ Tuesday, May 11, 8:30 - 10:00 Cache Memories Alan J. Smith CS Division, 571 Evans Hall UC Berkeley Berkeley, CA 94720 Abstract Cache memories are used in almost all high performance computer systems in order to effectively reduce the main memory access time. This talk will survey the design of and design considerations for cache memories. Specific topics discussed will include: cache fetch algorithms (demand vs. prefetch), placement (set associative, direct mapping, etc.) and replacement (LRU, FIFO, etc.) algorithms, line size, store through vs. copy back updating of main memory, cold start vs. warm start miss ratios, multicache consistency, the effect of input/output through the cache, virtual address caches, user/supervisor caches, multilevel cache, the behavior of split instruction/data caches, cache size, translation lookaside buffers, etc. ...................................................... Tuesday, May 11, 10:30 - 12:00 Issues in Trace-Driven Simulation David Kaeli IBM T. J. Watson Research Center Yorktown Heights, NY 10598 Abstract Considerable effort has been devoted to the development of accurate trace-driven simulation models of today's computer systems. Unfortunately many modelers do not carefully inspect the input to their models. The fact is that the output of any model is only as good as the input to that model. This tutorial covers the many aspects of the input traces used in trace-driven simulation. A review of the many trace types, trace-driven model types, tracing methodologies, and trace-length reduction techniques is presented. A key issue addressed in this tutorial will show how traces are frequently misused in a simulation. The goal is to equip the modeler with enough background so that he or she can be more critical of the content of traces used to drive their simulation models. ................................................................ Tuesday, May 11, 1:30 - 3:00 Performance and Reliability in Disk Arrays Garth Gibson Computer Science Department Carnegie Mellon 5000 Forbes Ave Pittsburgh, PA 15213 Abstract Recent advances in processor and memory technology have given rise to increases in computational performance that far outstrip increases in the performance of secondary storage technology. Coupled with emerging small-disk technology, disk arrays provide the cost, volume, and capacity of traditional disk subsystems but, by leveraging parallelism, many times their performance. This tutorial will describe the effect of various configurations of data striping on I/O performance. Unfortunately, arrays of small disks may have much higher failure rates than the single large disks they replace. Redundant Arrays of Inexpensive Disks (RAID) use simple redundancy schemes to provide high data reliability. This tutorial will describe models for the performance costs and reliability enhancements of various configurations of redundant disk arrays. .............................................................. Tuesday, May 11, 3:30 - 5:00 Architectures and Algorithms for Digital Multimedia On-Demand Servers P. Venkat Rangan Multimedia Laboratory Computer Science and Engineering Department University of California at San Diego La Jolla, CA 92093 Abstract Recent advances in digital audio and video processing technologies coupled with the availability of high bandwidth networks and large capacity storage devices is making it feasible to build multimedia on-demand servers over metropolitan-area networks. In this tutorial, we will develop architectures and algorithms for designing a high-performance, multimedia on-demand server capable of servicing a large number of clients simultaneously. We will address questions such as: (1) How should digital continuous media be stored on disks so as to ensure their continuous and synchronous retrieval? (2) What policies should a multimedia server employ to service multiple clients simultaneously without violating the continuity or synchrony requirements of any of the clients? We will evaluate the current state of art in these areas, and outline promising areas for future research. ............................................................ Tuesday, May 11, 8:30 - 10:00 Evaluation of Real-Time Fault-Tolerant Computing Systems Kang Shin University of Michigan Department of Electrical Engineering and Computer Science Abstract Real-time computing systems that commonly support such critical applications as computer-integrated manufacturing, drive-by-wire, and process control must not only be logically correct, but also meet the stringent timing constraints imposed by the applications. This course will cover the state-of-art means to accomplish this goal and discuss how to characterize and evaluate real-time fault-tolerant systems. It will focus on the basic concepts of real-time fault-tolerant systems, real-time performance evaluation, task and message scheduling. If time allows, we will also touch on real-time architectures and communications, operating systems and software tools, and fault tolerance. This course is intended for engineers and scientists who desire a quick introduction to the state-of-art techniques for distributed real-time systems. capturing and analyzing performance data on massively parallel systems, including potential perturbations, large data volumes, and correlation of data with source code. Based on this background, we describe a software toolkit approach whose design is based on the lessons learned from two previous generations of performance data analysis software. The toolkit contains a set of performance data transformation modules that can be interconnected in user-specified ways to form an acyclic, directed data analysis graph. Performance trace data are represented in a self-documenting stream format that includes internal definitions of data types, sizes, and names. ................................................................. Tuesday, May 11, 1:30 - 3:00 Performance Engineering of Client-Server Software Systems C.M. Woodside Department of Systems and Computer Engineering Carleton University, Ottawa, Canada Abstract A great many distributed applications have a client-server architecture, and many more are being planned, including transaction processing systems, network services, and computational services. Almost all systems based on remote procedure calls (RPC) have a client-server structure and distributed processing architecture such as DCE (Distributed Computing Environment) from the open System Foundation do too. The performance of these systems has an unsettling characteristic because of the blocking nature of almost all RPC, that the service times of the software server tasks are not constant, but increase with load. This makes it more necessary than normal to model the performance of the system, using a model that incorporates this effect. Further, this effect also can modify the location of bottlenecks, and create "software bottlenecks" which are difficult to diagnose in advance. The tutorial will describe a special category of models for these systems called "Rendezvous Networks", their application to modelling large-scale client-server systems and will highlight issues involving the propagation of bottlenecks, the diagnosis and correction of bottlenecks and the performance effects of the partitioning of functionality between servers in complex systems. An analytic modelling framework for these systems will be used to give insight into their behaviour. This is an applied tutorial directed mainly towards people who want to build models of real systems. The methods are intended to be usable by system designers who are not performance professionals. Some recent research results will be included, and some research issues will be raised. ................................................ Tuesday, May 11, 3:30 - 5:00 Modelling and Analysis of Transaction Processing Systems Philip Yu IBM T. J. Watson Research Center Yorktown Heights, NY 10598 Abstract The objective is to provide an analytic framework for analyzing the various design issues in transaction processing systems. The transaction response time depends on the level of data contention and hardware resource contention. A hierarchical modelling approach is presented which decomposes the problem into submodels and captures the interactions between the submodels through a fixed point iteration. Various approaches to concurrency control (CC) are examined. A general analytic modelling methodology is presented so that different CC schemes can be analyzed under a unified framework to better understand their performance trade-offs. The analysis captures the effect of skewed data accesses, different lock modes, and variable length transactions. We first consider a centralized system and then generalize the methodology to analyze distributed systems and coupled systems. The effect of data replication under different distributed CC schemes is analyzed. The implications of buffer coherency control policies are examined in a multi-node environment. .................................................................... Tuesday, May 11, 8:30 - 10:00 The Stochastic Complement, Decomposition and Aggregation in the Numerical Solution of Markov Models Richard Muntz Department of Computer Science University of California at Los Angeles Los Angeles, Ca 90024 Abstract Numerical solutions of Markov models are useful when a closed form solution is not available. Notwithstanding the advances in high performance computing, the ability to compute numerical solutions is limited by the size of the state space. To solve models of realistic complexity is often impossible unless there is special structure that can be exploited. This tutorial will examine several classes of special structure that, when applicable, can lead to efficient numerical solution methods. We will consider the use of aggregation, decomposition and the concept of the stochastic complement in direct solution methods, exact iterative solution methods and in the computation of bounds on the solution of Markov models. To illustrate the use of these methods we will present applications which include models of reliability, load balancing, queueing networks, and ATM switches. ................................................................ Tuesday, May 11, 10:30 - 12:00 Divide-and-Conquer Methods for Quasi-Birth-and-Death Processes Guy Latouche Universite Libre de Bruxelles Departement d'Informatique Campus Plaine --- CP 212 Boulevard du Triomphe 1050 Bruxelles Belgium Abstract The M/M/1 queue is a Markovian process with homogeneous transitions: the arrival and service rate do not depend on the state of the queue. Many models in computer performance evaluation appear as homogeneous Quasi-Birth-and-Death processes, where the state space is divided into subvectors called levels, and the transition rates do not depend on the level. This homogeneity property allows one to use divide-and-conquer methods to determine various quantities of interest: the stationary distribution, expected time to blocking (in the finite buffer case), the distribution of the maximum queue length during a busy period, etc. In this tutorial, we shall give a simple presentation of the methods and discuss a few examples. ...................................................... Tuesday, May 11, 1:30 - 3:00 TES Modeling of Autocorrelated Time Series With Applications to Source Modeling of Compressed Video Benjamin Melamed NEC USA, Inc. Princeton, NJ 08540 Abstract This tutorial overviews a new class of diverse models, called TES (Transform-Expand-Sample), for generating autocorrelated random sequences. The novel feature of TES is its ability to simultaneously capture both the empirical marginal distribution (histogram) and the empirical autocorrelation function (a measure of linear dependence), calculated from empirical data sets (field measurements). A visual interactive software package, called TEStool, designed to support TES modeling, will also be described and its use will be demonstrated. The efficacy of TES will be exhibited through case studies of source modeling for compressed (VBR) video, including DCT and MPEG. ..................................................... Tuesday, May 11, 3:30 - 5:00 Modeling Dynamic Behavior With Time Series Analysis Joe Hellerstein IBM T. J. Watson Research Center Yorktown Heights, NY 10598 Abstract The need to model dynamic behavior in computer systems arises in many contexts, such as characterizing the locality of file access patterns to evaluate the effectiveness of disk caching, evaluating the convergence properties of control algorithms, and identifying performance problems by their time serial behavior. This tutorial describes time series analysis (a statistical technique), and applies it to the performance analysis of computer systems, especially to queueing models. The autoregressive integrated moving average (ARIMA) model is discussed in detail, including the use of autocorrelation functions for model identification. The tutorial concludes with a case study that examines measurements taken from a mainframe computer system. ---------------------------------------------------- PEOPLE General Chair: Susan Owicki, Digital Equipment Corp. Program Chair: Dick Muntz, UCLA Conference Committee: David Finkel, Worchester Tech - Book Display Annop Gupta, Stanford University - Local Arrangements Rejeev Jog, Kubota Pacific - Proceedings Editor Seetha Lakshmi, IBM - Treasurer Michael Melliar_Smith, UCSB - Registration Louise Moser, UCSB - Poster Sessions Randolph Nelson, IBM - Tutorials Herb Schwetman, MCC - Publicity Program Committee: Ian F. Akyildiz, Georgia Tech Peter Danzig, USC Larry Dowdy, Vanderbilt Derek Eager, Univ. of Saskatchewan Garth Gibson, CMU Albert Greenberg, AT&T Bell Labs Anoop Gupta, Stanford Raj Jain, DEC Benjamin Melamed, NEC USA Robert J.T. Morris, IBM Dave Nicol, William and Mary Dan Reed, Univ. of Illinois Scott Shenker, Xerox Edmundo de Souza e Silva, UFRJ Brazil Guri Sohi, Univ. of Wisconsin Billy Stewart, NCSU Don Towsley, Univ. of Massachusetts Johnny Wong, Univ. of Waterloo Philip S. Yu, IBM John Zahorjan, Univ. of Washington Songnian Zhou, Univ. of Toronta Invited Speakers: Lennart Johnsson, Thinking Machines Corp. and Harvard University David Patterson, University of California at Berkeley --------------------------------------------------- LOCATION Santa Clara and the San Francisco Bay Area Santa Clara is located in the heart of Silicon Valley, just 45 minutes South of San Francisco and 10 minutes North of San Jose. The conference site is adjacent to Great America Theme Park (open on week ends in May) and near wineries, Spanish missions, and the Santa Cruz beach and boardwalk. You can extend your stay in California with a weekend in San Francisco, Or tour the wine country North of San Francisco, or enjoy the charm and natural beauty of the Monterey Bay to the South. All are within easy driving distance of the conference site. Climate The weather in Santa Clara in May is typically sunny and mild. Days are sometimes hot, and a light jacket may be neeeded in the early morning and evening. Conference Site All technical sessions, lunches, and registration will be held at the Santa Clara Marriott. The hotel has free parking, indoor/outdoor pool, whirlpool, exercise room, game room, and four lighted tennis courts. Golf is available nearby. Directions >From San Jose Airport: Go four miles north on Highway 101 to Great America Parkway Exit. From San Francisco International Airport: Go 36 miles south on Highway 101 to Great America Parkway Exit. The Marriot is at the intersection of Great America Parkway and Mission College Boulevard, a short distance from Highway 101. Transportation Complimentary van transportation to the Marriott is available from San Jose Airport. Shuttle service from San Francisco International Airport is offered by the following companies. Advance reservations are required. Express Shuttle $21 408 378-6720 BayPorter $17 415 467-1800 Airport Connection $16 4l5 363-1500 -------------------------------------------------------------- CONFERENCE AND TUTORIAL PRE-REGISTRATION Please send with check or money order to: SIGMETRICS '93 Registration c/o Prof. P. M. Melliar-Smith Electrical and Computer Engineering University of California Santa Barbara, CA 93106 Telephone: 805-893-8438 Fax: 805-893-3262 Email: pmms@ece.ucsb.edu Attendance options (circle one): Before April 12 After April 12 Tutorials & Conference (May 10-14) ACM/SIG Member $460 $610 Non-Member $535 $685 Full-time Student $180 $260 Conference Only (May 12-14) ACM/SIG Member $310 $385 Non-Member $360 $435 Full-time Student $115 $165 Tutorials Only (May 10-11) ACM/SIG Member $210 $285 Non-Member $235 $310 Full-time Student $ 80 $115 Vegetarian meals requested:____________ Extra Banquet Ticket (May 13) $45__________ Total Amount Enclosed:_____________________________ If you have selected the student rate, please be prepared to show a valid student ID at the time you attend the conference. The conference registration fee includes attendance at all technical sessions, two luncheons, reception and banquet, and a copy of the proceedings. The tutorial registration fee includes attendance at all tutorial sessions, tutorial materials, and one luncheon. In the event that you are unable to attend the meeting after you have paid fees, you will receive full refund prior to April 28, 75% refund from April 18 to May 7, and no refunds after May 7. Name:______________________________________________________ Phone number:______________________________________________ Affiliation:_________________________ Mbr number:__________ Address:___________________________________________________ City:_____________________State:____________Zip:___________ E-mail address:____________________________________________ ____You may include my name in the List of Participants which might be distributed at the Conference. A FREE Conference T-shirt will be available to those who pre-register and the pre-registration is received by the due date, April 12. Please indicate size and color: _____Medium _____Large _____X-large ____Burgandy _____While ------------------------------------------------------- HOTEL RESERVATION - ACM SIGMETRICS '93, May 9 - 15, 1993 Send to: Santa Clara Marriot Hotel 2700 Mission College Boulevard Santa Clara, CA 95054 Tel: 408-988-1500, Fax: 408-727-4353 Deadline for conference room rate: April 17, 1993 $95.00 Single Double Smoking $105.00 Triple Quadruple Nonsmoking plus Salse and lodging tax (9.56%) Arrival Date:____________________ Departure data:________________ Arrival Time:____________________ All reservations will be held until 6:00 pm on arrival night unless accompanied by a first night room deposit on a major credit card. If the room is guaranteed and the individual does not arrive, the individual will be responsible for the first night payment of room rate plus tax charges. Card type:__________________, Expiration date:_________________ Card number:____________________ Signature:________________________________ Note: Check in time is after 3:00 pm and check out time is at 11:00 am. Reservations received after April 17th will be confirmed only if space is available. Guests who cancel by 6:00 pm on scheduled arrival will not be charged. Cash, major credit card and checks are acceptable when settling your bill. Do you wish confirmation to be sent to you? yes________, no__________ Name:_________________________________________________ Company/Institute:____________________________________ Address:______________________________________________ City & State/Prov:____________________________________ Country:________________ Zip/Postal code:_____________ Telephone:____________________________________________ ------------------------------ END OF SIMULATION DIGEST ************************