NetNews Usenet Archive 1992 #16

home *** CD-ROM | disk | FTP | other *** search

/ NetNews Usenet Archive 1992 #16 / NN_1992_16.iso / spool / comp / parallel / 1809 < prev next >

Wrap

Text File | 1992-07-27 | 21.9 KB | 577 lines

Newsgroups: comp.parallel Path: sparky!uunet!darwin.sura.net!gatech!hubcap!fpst From: Jane Hillston <jeh@dcs.ed.ac.uk> Subject: Call for Participation: Performance Techniques and Tools Conference Message-ID: <1992Jul27.141151.27157@hubcap.clemson.edu> Sender: nnews@dcs.ed.ac.uk Organization: Edinburgh University Date: 27 Jul 92 09:31:00 GMT Approved: parallel@hubcap.clemson.edu Lines: 565 CALL FOR PARTICIPATION 6TH INTERNATIONAL CONFERENCE ON MODELLING TECHNIQUES AND TOOLS FOR COMPUTER PERFORMANCE EVALUATION September 16-18th, 1992 Scandic Crown Hotel Edinburgh, Scotland, UK. The sixth International Conference on Modelling Techniques and Tools for Computer Performance Evaluation will be held in Edinburgh, UK. This is the major European forum for the presentation of research and applications of performance analysis tools and techniques. The conference will be held in the Scandic Crown Hotel, a modern luxury hotel in the middle of Edinburgh's historic Old Town, on the Royal Mile. The provisional programme is given below: WEDNESDAY 16th SEPTEMBER 9:20 - 9:30 Opening Remarks 9:30 - 10:30 Arnold Allen - invited paper Session Chairman: R. Pooley 10:30 - 11:00 Coffee Break 11:00 - 12:00 PETRI NETS (Chairman: G. Balbo) C. Lindemann DSPNexpress: A Software Package for Efficiently Solving Deterministic and Stochastic Petri Nets J.A. Rolia and K.C. Sevcik Fast Performance Estimates for a Class of Generalized Stochastic Petri Nets 12:00 - 12:30 Overview of tools to be demonstrated (Chairman: M. Calzarossa) 12:30 - 2:00 Lunch and Tool Demonstrations 2:00 - 3:00 MEASUREMENTS (Chairman: D. Thomas) J.K. Flanagan, K. Grimsrud et al BACH: BYU Address Collection Hardware N.J. Gunther On the Application of Barycentric Coordinates to the Prompt and Visually Efficient Display of Multiprocessor Performance Data 3:00 - 3:30 Tea Break 3:30 - 5:30 WORKLOADS (Chairman: G. Serazzi) B.M. Carlson et al Speedup Properties of Phases in the Execution Profile of Distributed Parallel Programs K.J. McDonell Benchmark Frameworks and Tools for Modelling the Workload Profile M.K. Acharya et al. Real-time Hierarchical Traffic Characterization of a Campus Area Network A. Nanda and L.M. Ni Impact of Memory Contention and Synchronisation on Multiprocessor Performance Measurement and Calibration THURSDAY 17th SEPTEMBER 9:00 - 10:30 ESPRIT III projects - invited papers Chairman: P. Hughes Trevor Chambers - ESPRIT project 6942 PEPS: Performance Evaluation of Parallel Systems Horst Lindmeier - ESPRIT project 6271 Benchmarking for Embedded Control and Real-time Applications Kam-Fai Wong - ESPRIT project 7091 Pythagoras: Performance Quality Assessment of Advanced Database Systems 10:30 - 11:00 Coffee Break 11:00 - 12:30 METHODS I (Chairman: R. Puigjaner) E.W. Brehm, R.T. Goettge and F. McCaleb START/ES - An Expert System Tool for System Performance and Reliability Analysis A. Valderruten et al. Deriving Queueing Networks Performance Models from Annotated LOTOS Specifications J. Hillston A Tool to Enhance Model Exploitation 12:30 - 2:00 Lunch and Tool Demonstrations 2:00 - 4:00 PARALLEL AND DISTRIBUTED SYSTEMS (Chairman: R. Candlin) F. Hartleb and V. Mertsiotakis Bounds for the Mean Runtime of Parallel Programs G. Lyon, R. Snelick and R. Kacker Time-Perturbation Tuning of MIMD Programs A. Birman, D. Ferguson and Y. Kogan Asymptotic Solutions for a Model of Large Multiprocessing Systems with Multiple Workloads R. Kroeger, M. Gergeleit and F. Lange JEWEL: Experiences with a Distributed Measurement System 4:00 - 5:15 Tea Break and Poster Session (Poster Session Organiser: J. Hillston) FRIDAY 18th SEPTEMBER 9:00 - 10:00 Raj Jain - invited paper Chairman: R. Marie 10:00 - 11:00 METHODS II (Chairman: G. Haring) A.L. Opdahl Sensitivity Analysis of Combined Software and Hardware Performance Models: Open Queueing Networks N.M. Patel Structuring Analytical Performance Models using Mathematica 11:00 - 11:30 Coffee Break 11:30 - 12:30 METHODS III (Chairman: P. Harrison) R. Chakka and I. Mitrani A Numerical Solution Method for Multiprocessor Systems with General Breakdowns and Repairs P. Buchholz Hierarchical Markovian Models - Symmetries and Reduction 12:30 - 2:00 Lunch and Tool Demonstrations 2:00 - 3:30 CASE STUDIES (Chairman: A. Wight) A.R. Hajare Multiprocessor Performance Evaluation: A Case Study of Two Modelling Tools S. Salza and R. Tomasso A Modelling Tool for the Performance Analysis of Relational Database Applications M. Malhotra and K.S. Trivedi Reliability of Redundant Arrays of Inexpensive Disks (RAID) 3:30 - 4:00 Tea Break 4:00 - 4:15 Closing Remarks POSTERS AND TOOL DEMONSTRATIONS A.E. Krzesinski and N.J. Townsend /DASD: A Tool to Model the Performance of IBM DASD Systems X. Zhang, N. Nalluri and X. Qin MIN-Graph: A Tool for Monitoring and Visualising MIN-based Multiprocessor Performance B.R. Haverkort, A.P.A. van Moorsel and A. Dijkstra MGMtool: A Performance Modelling Tool Based on Matrix Geometric Techniques A. Hutchison and P.S. Kritzinger XWAN: A Tool to Model Adaptive Routing in a Wide Area Network A.E. Kostin Microsim: E-Net Tool for Modelling Concurrent and Distributed Data Processing Systems H. Heiss and A. Payer PASTE: A Tool for Evaluation of Processor Allocation Strategies G. Fleischmann and M. Gente Modelling and Evaluation of Parallel Programs using GIANT G. Fairhurst and A.Z.M. Salleh Modelling the Performance of the G.764 Packetised Voice Protocol R.L. Klevans and W.J. Stewart XMARCA: An X-Windows Interface for Queueing Network Modelling G. Chiola, R. Gaeta and M. Ribaudo Designing an Efficient Tool for Stochastic Well-Formed Coloured Petri Nets A.J. Field and R.L. While A General Framework for Discrete-Event Simulation Using Functional Languages TUTORIALS Before the conference, on Tuesday 15th September there will be a programme of tutorials on aspects of performance modelling. There will be two long tutorials in the morning and two sets of two short tutorials in the afternoon. Delegates may register for either a full day or a half day programme. Please indicate choice of tutorial(s) on the registration form. 9:00 - 12:00 A Nico M. van Dijk, University of Amsterdam, The Netherlands The Product Form Tool B Kishor S. Trivedi Duke University, USA Techniques and Tools for Reliability Modelling 12:00 - 1:30 LUNCH 1:30 - 3:00 C Ian F. Akyildiz Georgia Institute of Technology, USA Computational Techniques D Edmundo de Sousa e Silva University of Rio de Janeiro, Brazil Numerical Approaches 3:00 - 3:30 COFFEE 3:30 - 5:00 E Gabriele Kotsis University of Vienna, Austria Visualization of Performance Information F Boudewijn R. Haverkort University of Twente, The Netherlands Performability Modelling Tools Abstracts for each tutorial are given below: A THE PRODUCT FORM TOOL - Nico M. van Dijk Queueing network modelling has become a popular tool in telecommunication, computer performance evaluation and flexible manufacturing. The tutorial aims to highlight an approach that can be used as an engineering tool to evaluate stochastic service or computer networks. To this end, it will cover the following two topics: 1) Product forms and partial balance 2) A simple bounding methodology. Part 1 will discuss the physical notion of station balance as a key-property to deduce closed product form expressions for queueing networks with blocking. In concrete situations this notion will enable one to directly deduce whether or not a system will have a product form solution, and if not, why not. In part 2 these insights will be used further to also obtain simple performance bounds for non-product form queueing networks, most notably networks with blocking. The steps involved will be outlined and illustrated for some typical non-product form examples such as: - an overflow model - a system with breakdowns - finite tandem or assembly lines B TECHNIQUES AND TOOLS FOR FOR RELIABILITY MODELLING - Kishor S. Trivedi Various Markov reward models are described and investigated as a tool to evaluate reliability and availability models for computer networks, for example fault-tolerant systems. Both transient and steady-state measures are considered. Techniques covered are: - differential equations - numerical solution procedures - uniformization techniques and truncations - simulation procedures and enhancements Particularly, a hierarchical modelling tool is presented, called SHARPE (Symbolic Hierarchical Automated Reliability and Performance Evaluation), that combines the different techniques and models. This tool provides a specification language and solution methods for a large number of models such as: - series-parallel reliability block diagrams - reliability graphs and fault trees - cyclic and acyclic Markov chains - Generalized Stochastic Petri Nets - PMS (Processor-Memory-Switch) Models C COMPUTATIONAL TECHNIQUES FOR COMPUTER NETWORKS - Ian F. Akyildiz Much of the success of queueing network modelling for computer networks can be attributed to: - the existence of Jackson's closed product form expression and many extensions thereof, and - efficient computational techniques to actually compute these forms or related performance measures such as the throughput, sojourn time or loss probability. This tutorial will survey a number of these computational techniques ranging from the classical convolution algorithm to various exact and approximative versions of the MVA (mean value analysis) method like LBANC and LBANCBLO. In doing so, special attention is given to different blocking protocols when finite capacities are involved. Finally some recent results are presented to transform networks with blocking into networks without blocking. D NUMERICAL APPROACHES TO PERFORMABILITY - Edmundo de Souza e Silva The performance of computer networks under different loads and circumstances is a relevant indication of its being well-designed and operable. Computer systems reliability and availability deals with the representation of changes in structure of the system due to faults or breakdowns. As analytic tools are often not feasible to capture such aspects while simulation techniques may be unreliable, numerical possibilities are still searched for. An overview is provided of past work on performability evaluation by means of Markov chains and how relevant performance measures can be computed efficiently by numerical approaches. Particularly two important topics are highlighted: - the uniformization or randomization technique to transform continuous-time models into discrete-time models - the state space explosion by which actual computation becomes prohibitively expensive. Several techniques to deal with this problem are discussed. E VISUALIZATION OF PERFORMABILITY INFORMATION - Gabriele Kotsis The art of visualizing the performance of computer systems has gained increasing interest in the past few years. The pictorial representation of performance data will help the system analyst to obtain a better understanding of what is going on in the system, how it works and why it works (or why not). Visualized information makes use of the eye-brain system, with its great pattern recognition capability, in a way that is impossible with purely numeric data. This tutorial aims to report on the state of the art in performance visualization from the analyst's point of view. The usefulness of visualization techniques and tools supporting performance analysis will be investigated. In particular, the tutorial will contain three modules: The State of the Art in Performance Visualization This presentation will include: visualization primitives and techniques, static versus dynamic displays, display of univariate and multivariate data, categorization and comparison of existing visualization tools. Representing Performance Characteristics of Conventional Computer Systems The following topics will be discussed: visualization of hardware and software performance information, meaningful metaphors and easy recognizable presentation techniques. Visualizing the Performance of Parallel and Distributed Systems The particular problems arising in visualizing the behaviour of parallel systems (representations for massive amounts of data, representation of concurrency, visualization of hardware-software mapping) will be demonstrated along with possible solution techniques. F PERFORMABILITY MODELLING TOOLS - Boudewijn R. Haverkort Over the last few years Markov reward models (MRMs) have been used extensively for the analysis of the performability of computer and communication systems. Using MRMs for this purpose brings along two basic problems: model generation and model solution. Model solution has been the subject of study in a large number of papers. Far less attention has been paid to the model generation problem. In this tutorial the mathematical background of MRMs for performability modelling is briefly addressed. The emphasis is then on high-level specification techniques for performability models and various generation techniques that can be used to automatically generate MRMs from these specifications. Aspects addressed are modelling requirements (coming from the various application areas), high-level specification techniques (such as stochastic Petri nets, stochastic activity nets, dynamic queueing networks and object-oriented techniques), implementability of conceptual models, implementation aspects of MRM generation techniques and existing software tools for performability modelling and analysis. SOCIAL PROGRAMME Tuesday, 15th September 7.00pm Civic Reception at the Edinburgh City Chambers Wednesday, 16th September - Optional Excursions Murder and Mystery Tour Whisky Tasting Thursday, 17th September 7.00pm for 7.30pm Conference Dinner at the Royal College of Physicians -------- PROGRAMME COMMITTEE: Ian Akyildiz, Georgia Tech, USA Heinz Beilner, University of Dortmund, Germany Maria Calzarossa, University of Pavia, Italy Rosemary Candlin, University of Edinburgh, UK Giovanni Chiola, University of Torino, Italy Blaine Gaither, Amdahl Inc, USA Gunther Haring, University of Vienna, Austria Peter Harrison, Imperial College, London, UK Peter Hughes, Bell Northern Research (Europe), UK Peter King, Heriot-Watt University, UK Anthony Krzesinski, University of Stellenbosch, RSA Raymond Marie, IRISA, France Ramon Puijganer, University of Balearics, Spain Herb Schwetman, MCC Inc, USA Guiseppe Serazzi, Politecnico di Milano, Italy Connie Smith, L & S Computer Technology, USA David Thomas, BT Ltd, UK Alex Wight, University of Edinburgh, UK Organisation: Local Arrangements: Jane Hillston, (jeh@dcs.ed.ac.uk) Dept. of Computer Science, Univ. of Edinburgh, UK Sponsorship & Exhibitions: Peter King, (pjbk@dcs.hw.ac.uk) Dept. of Computer Science, Heriot-Watt Univ., UK Tutorials: Nico van Dijk, (fax: +31 20 525 4217) Dept. of Econometrics, Univ. of Amsterdam, Netherlands. -------- Travel Arrangements Air Edinburgh International Airport is less than 20 minutes from the city center and provides access to a worldwide network of international flights. There are 26 return flights each day to London provided by British Airways, British Midland and Air UK (Gatwick), in addition to regular services to Birmingham and Manchester. Direct air links to New York, Chicago and Boston, together with many main European cities are provided from Glasgow Airport, one hour's drive from Edinburgh. Rail Edinburgh benefits from an extensive network of rail services from London and other principal UK cities. High speed trains operate on the London route with a journey time of 4 hours to Waverley Station, in the heart of the city, a short walk from the Scandic Crown hotel. Inter City Sleeper services are provided on most routes. Road Express coach services are available to and from all major UK cities, operating on a comprehensive motorway or trunk road system. -------- The conference fee includes attendance at all sessions of the conference; refreshments and lunch each day; a local copy of the proceedings; attendance at the civic reception and the conference dinner; car parking at the hotel. The student conference fee does not include the conference dinner. Additional tickets for the conference dinner, price 35 pounds, will be available on arrival. Delegates registering before 14th August can take advantage of a preferential rate. Cancellation before 14th August will be eligible for 100% refund of all fees. Subsequent cancellation, up to 4th September, will be eligible for 75% refund. After that date no refund will be possible. Delegate substitution will be acceptable at any time. To register please detach and complete the form below and return it to: Mrs Edith Field UNIVED Technologies Ltd. University of Edinburgh, 16, Buccleuch Place, Edinburgh EH8 9LN SCOTLAND. efield@castle.ed.ac.uk fax: +44 31 650 6532 8<---8<---8<---8<---8<--- please detach and print 8<---8<---8<---8<--- Name ____________________________________________ Company/Institute________________________________ Address__________________________________________ _________________________________________________ _________________________________________________ _________________________________________________ Telephone________________________________________ Email____________________________________________ Signed___________________________ Date___________ CONFERENCE REGISTRATION Before 14/8/92 After 14/8/92 Academic Delegate 275 pounds 295 pounds ______ pounds Non-academic Delegate 325 pounds 345 pounds ______ pounds Student Delegate 150 pounds 170 pounds ______ pounds (Students will be required to bring some form of identification of certification of student status.) TUTORIAL REGISTRATION Which tutorial(s) would you like to attend? (Please tick) A ___ B ___ C ___ D ___ E ___ F ___ Tutorial fee includes lunch and refreshments during the day, tutorial notes of the tutorial attended and attendance at the civic reception. Half day tutorial attendance 50 pounds Full day tutorial attendance 80 pounds ______ pounds Do you require vegetarian meals? Yes/No ______ Accommodation has been reserved in the Scandic Crown Hotel and in Pollock Halls, the University's major residence complex in the south of the city, approximately 20 minutes walk from the Scandic Crown. Scandic Crown Hotel: 64.75 pounds per person per night (B&B) single or double room _________ pounds 59.00 pounds per night for accompanying spouse sharing double room _________ pounds Pollock Halls: 21.75 pounds per person per night (B&B) (single rooms) _________ pounds Arrival Date _____________ Departure Date _____________ TOTAL COST __________ pounds Cheques, Money Orders and Bank Drafts in Sterling payable to UNIVED Technologies Ltd.