Very Computer

by **Jerome Rol** » Sat, 08 Apr 1995 04:00:00

******************************************************************************
SIGMETRICS/PERFORMANCE '95 ADVANCE PROGRAM and PRE-REGISTRATION

CALL FOR PARTICIPATION

!!! ! $$$ LAST CHANCE TO REGISTER AT LOW PRE-REGISTRATION PRICES $$$ ! !!!

Note: We recommend that you reserve your hotel room early.

******************************************************************************

1995 ACM SIGMETRICS/PERFORMANCE '95
Joint
International Conference on
Measurement and Modeling of Computer Systems

Sponsors: ACM, IFIP W.G. 7.3

Date: May 15 - 19, 1995

Place: The Westin Hotel, Ottawa, Ontario, Canada
==============================================================================
OVERVIEW

The 1995 joint conference of PERFORMANCE '95 and ACM SIGMETRICS '95
will be held May 15 - 19, 1995, in Ottawa, Canada. ACM SIGMETRICS
is an anual conference and PERFORMANCE occurs overy 18 months.
Traditionally, a joint conference is held every third year. This joint
conference features:

Tutorials (May 15, 16)
Technical Conference (May 17-19) with:
Poster and Hot Topic sessions
Presentation of papers

This on-line announcement contains the following sections:

LOCATION
PRE-REGISTRATION
HOTEL RESERVATION
TUTORIALS (May 15-16)
TECHNICAL PROGRAM (May 17-19)
COMMITTEES
SPONSORS

==============================================================================
LOCATION

Ottawa, Ontario, Canada

Ottawa, Canada's capital city, is home to many unique Canadian institutions,
fine museums, and monuments of national significance. Ottawa is set against
the backdrop of flowing rivers and a canal. There is an abundance of green
space which gives Ottawa a parkland atmosphere unlike any other capital city
in the world.

In mid-May, the city is brightened by a brilliant display of over 1,000,000
tulips. Princess (later Queen) Juliana of the Netherlands and her family
took refuge in Ottawa during World War II. When the family returned home,
they presented Ottawa with a huge gift of bulbs. Since then, more than 10,000
bulbs have arrived every year.

The conference hotel is centrally located, next to the Rideau Canal and the
Byward Market. The market is 150 years old and is a center for boutiques,
restaurants, and entertainment. The hotel is within walking distance of many
fine restaurants, good shopping, the spectacular Parliament Buildings, and
many of Ottawa's museums and other attractions. There are many other hotels
near by including the Lord Elgin, and Novotel.

Also during the months of May and June, Ottawa will host a celebration of
the 50th anniversary of the end of the Second World War. Because of the
unpredictable effect of this latter event on hotel space, you are urged to
book your hotel room early!

Climate

The weather in Ottawa in May is variable and can be beautiful. Daytime
temperatures between 50F-70F (12-20C) are typical, with temperatures as
low as 40F (5C) in the early morning and evening.

Exchange Rate

At the time of this message, one Canadian dollar $1CA = $0.71US.

Conference Site

All technical sessions, lunches, and registration will be held at
the Westin Hotel. The alternative hotels are within walking distance
of the conference hotel and are listed below.

Air Travel

From the USA, direct flights to Ottawa's MacDonald-Cartier International
Airport are available from Baltimore, Newark, La Guardia, Boston, Pittsburgh,
Syracuse and Washington. Most other flights from the USA, Europe, and from
Pacific rim countries require a change at Pearson International Airport in
Toronto or Mirabel International Airport in Montreal. If you have
a choice come via Toronto since there are frequent connecting flights
to Ottawa. There are direct flights to Ottawa from London England,
Amsterdam, and Paris.

Transportation

A taxi from the MacDonald-Cartier Airport to the downtown hotels
is approximately CA$20. An airport bus is also available.

Many flights from Europe arrive at Montreal's Mirabel Airport.
There are buses and connecting flights from Mirabel to Ottawa,
both infrequent. Mirabel is approximately 150km from Ottawa,
so renting a car is feasible.

For those renting a car at Ottawa's MacDonald-Cartier International Airport:
follow the airport parkway to Colonel By Drive. Turn right (east) on Colonel
By Drive. The Westin is located just before the intersection of Colonel By
Drive and Rideau Street (about 3 miles (5km) from the right turn).

==============================================================================
PRE-REGISTRATION FORM

For your convenience, payment may be made in either United States (US)
or Canadian (CA) dollars with an international money order, or cheque
drawn on an American or Canadian bank. Please send cheque or money
order made payable to:

SIGMETRICS '95 Registration
c/o Conference Coll Inc.
1138 Sherman Drive
Ottawa, Ontario, Canada, K2C 2M4

For further information:
Tel: 1-613-224-1741
Fax: 1-613-224-9685
Email: c...@sce.carleton.ca

You may fax in your registration and follow up with payment
by mail. Confirmation will be sent only upon receipt of
payment.

------------------------------------------------------
******The following rates are in US Dollars US$*******
------------------------------------------------------
Attendance options (circle one): Before April 15 After April 15
Tutorials & Conference (May 15-16)
ACM/SIG/or IFIP WG7.3 Member $460 $620
Non-Member $545 $695
Full-time Student $185 $270

Conference Only (May 17-19)
ACM/SIG/or IFIP WG7.3 Member $310 $385
Non-Member $360 $435
Full-time Student $105 $155

Tutorials Only (May 15-16)
ACM/SIG/or IFIP WG7.3 Member $205 $285
Non-Member $230 $310
Full-time Student $85 $120

Special Dietary meals requested (indicate request):____________

Extra Banquet Ticket (May 18) (48US$):___________

Total Amount Enclosed (US$):_____________________________
-----------------
End US Funds Part
-----------------

------------------------------------------------------------
******The following rates are in Canadian dollars CA$*******
------------------------------------------------------------
Attendance options (circle one): Before April 15 After April 15
Tutorials & Conference (May 15-16)
ACM/SIG/or IFIP WG7.3 Member $612 $873
Non-Member $725 $979
Full-time Student $246 $380

Conference Only (May 17-19)
ACM/SIG/or IFIP WG7.3 Member $412 $542
Non-Member $479 $613
Full-time Student $140 $218

Tutorials Only (May 15-16)
ACM/SIG/or IFIP WG7.3 Member $273 $401
Non-Member $306 $437
Full-time Student $113 $169

Special Dietary meals requested (indicate request):____________

Extra Banquet Ticket (May 18) (60CA$):___________

Total Amount Enclosed (CA$):_____________________________
-----------------------
End Canadian Funds Part
-----------------------

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, three 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. Student registration
does not include a banquet ticket.

In the event that you are unable to attend the meeting after you have paid
fees, you will receive full refund prior to April 15. No refunds will be
available after April 15 but you will be sent a copy of the proceedings.

(Please print)

First Name:______________________________Last Name:________________________

Affiliation:_________________________

Address:___________________________________________________

City:_____________________Prov/State:____________Postal/Zip:___________

Phone number:______________________Fax Number:________________________

E-mail address:____________________________________________

Membership Number:______________________
* For ACM members give your ACM membership number, for
IFIP members use "IFIP", for students please indicate
"student" followed by your student number

You may include my name in the List of Participants to be
distributed at the Conference:

Yes______ No______

Signature _____________________

END OF PRE-REGISTRATION FORM

==============================================================================
HOTEL RESERVATION - ACM SIGMETRICS/PERFORMANCE '95, May 14 - 19, 1995

Please phone or fax your reservation request directly to one of following
hotels and request a confirmation number. Ask for the room block
under the name sig95. Non-smoking rooms are available upon request.

Conference Hotel:
----------------

The Westin Hotel
11 Colonel By Drive
Ottawa, Canada, K1N 9H4
Phone: 1-613-560-7000
Fax: 1-613-234-5396
Toll Free in U.S.A. and Canada: 1-800-228-3000

Rate: $115CA/night for single or double room, plus 12% taxes.

Deadline for conference room rate: April 12

Check-in time 3pm, Check-out time 1pm

Alternate Hotels:
-----------------

The Lord Elgin Hotel
100 Elgin Blvd.
Ottawa, Canada, K1P 5K8
Phone: 1-613-235-3333
Fax: 1-613-235-3223
Toll Free in U.S.A. and Canada: 1-800-267-4298

Rate: $89CA/night for single or double room, plus taxes.

Five minute walk to the Westin Hotel.

Deadline for conference room rate: April 12

Novotel
33 Nicholas St.
Ottawa, Canada, K1N 9M7
Phone: 1-613-230-3033
Fax: 1-613-230-4186
Toll Free in U.S.A. and Canada: 1-800-668-6835

Rate: $95CA/night for single or double room, plus taxes.

Two minute walk to the Westin Hotel.

Deadline for conference room rate: April 12

========================================================================
COMMITTEES

Conference Committee

General Chair: Murray Woodside (Carleton University):

Larry Dowdy (Vanderbilt) - Tutorials
David Finkel (Worcester) - Proceedings
Gabriele Kotsis (Vienna) - Publicity (Europe)
Fred Kaudel (BNR) - Finance
Shikharesh Majumdar (Carleton) - Local Arrangements
Isi Mitrani (Newcastle) - Program
Dorina Petriu (Carleton) - Registration
K. K. Ramakrishnan (DEC) - Book Display
Jerome Rolia (Carleton) - Publicity (Americas-Other)
Edmundo de Souza e Silva (Rio de Janeiro)- Awards
Satish Tripathi (Maryland) - Program

Program Committee

Program Co-Chairs: Satish Tripathi (USA) and Isi Mitrani (UK):

Ashok Agrawala (USA) Tom Anderson (USA)
Francois Baccelli (France) Onno Boxma (Netherland)
Rick Bunt (Canada) Ed Coffman (USA)
Domenico Ferrari (USA) Erol Gelenbe (USA)
Albert Greenberg (USA) Gunter Haring (Austria)
Phil Heidelberger (USA) Bill Henderson (Australia)
Ulrich Herzog (Germany) Mark Hill (USA)
Giuseppe Iazeolla (Italy) Jim Kurose (USA)
T. V. Lakshman (USA) Ed Lazowska (USA)
Danny Menasce (USA) Debasis Mitra (USA)
Dick Muntz (USA) Harry Perros (USA)
Brigitte Plateau (France) S. V. Raghavan (India)
Dan Reed (USA) Martin Reiser (Switzerland)
Giuseppe Serazzi (Italy) Ken Sevcik (Canada)
Kishor Trivedi (USA) Mary Vernon (USA)
John Zahorjan (USA)

==============================================================================
LOCAL SPONSORS
Bell Northern Research (BNR) - Ottawa
IBM Canada Center for Advanced Studies - Toronto
National Research Council of Canada (NRC) - Ottawa
===============================================================================
TUTORIALS, MAY 15-16 1995

Tutorial Slate at a Glance
SIGMETRICS/PERFORMANCE'95

Tutorial Slate, detailed descriptions
-------------------------------------

Monday, May 15, 1:30 - 5:00 (Introductory Track)

Performance Evaluation 101: Basic Modeling Techniques

Larry Dowdy
Vanderbilt University

This tutorial is aimed at novice performance modelers. Very little
background knowledge is assumed. The context and motivating examples
will focus on performance prediction. Given a computer system with
several options for possible improvement, a performance prediction model
can be used to evaluate each option. The option with the best predicted
performance, relative to the implementation cost, is judged to be the
best alternative. Topics include: measurement, workload characterization,
model construction, model solution, calibration, prediction, and
validation. Each of these topics involves difficult problems that are
often ignored. However, the practitioner cannot ignore these problems.
The presentation will be example oriented and intuition driven. Basic
modeling techniques using Markovian analysis, convolution, mean value
analysis, Petri net models, bounding techniques, decomposition, and
approximate analysis will be covered. Active participation by the
audience will be required via self-assessment exercises.

******************************************************************************
Monday, May 15, 1:30 - 3:00 (Intermediate Track)

ATM Local Area Networks

Dipak Ghosal
Bellcore

ATM has emerged as the leading technology which has the potential to
remove the performance bottlenecks in todays LANs. Along with the promise
comes the challenge to gracefully interwork ATM with the huge embedded
base of legacy LAN protocols and architectures. In this tutorial we will
address the various design and performance issues in interworking ATM
networks with the legacy LANs. We will study two proposals, namely, the
the LAN Emulation drafted by the ATM Forum and the classical IP over ATM
drafted by the IETF. We will discuss the relevant routing and addressing
issues and review the signalling protocols and architectures that are
needed to transport legacy LAN traffic over ATM networks. Finally, we
will identify some of the key network and traffic management issues in
ATM local area networks.

******************************************************************************
Monday, May 15, 1:30 - 3:00 (Advanced Track)

Parallel Complexity and Real Life Parallel Performance

David Nicol
College of William and Mary

This tutorial is a survey in the notions used to talk about the parallel
complexity of algorithms. It is intended for anyone who is mathematically
literate, but who is unfamiliar with complexity analysis. The tutorial
begins with a review of the tools one uses in complexity analysis, leading
up to the idea of problem transformations. The relationship between the
time required to solve a problem and the space required is discussed,
which leads us to the notion of P-complete problems. The practical
implications of P-completeness are discussed. Finally, the complexity of
some real life problems important to parallel processing are discussed in
the light of the introductory material.

******************************************************************************
Monday, May 15, 3:30 - 5:00 (Intermediate Track)

Design and Implementation of Distributed,
Fault Tolerant, On-Time Systems

Ashok K. Agrawala
University of Maryland

On-time systems are those in which there are tasks which must be executed
at specific time instants (i.e., executed with tight jitter constraints).
Such requirements appear in a variety of systems with different degrees
of tight timing constraints. In this tutorial we address the design and
implementation issues for such systems when they must operate in a
distributed environment and must support fault tolerance requirements.
The knowledge of the resource requirements of an application is essential
in its on-time execution. In order to meet the temporal constraints of
on-time applications, the system must support suitable resource management
mechanisms. Depending on the characteristics of the application,
preemption may or may not be permitted. Priority based techniques, such
as rate monotonic, may be used for the support of hard real-time execution
of applications with preemption. While rate monotonic scheduling may meet
the hard real-time constraints, it may not be able to meet the jitter
constraints of on-time execution. Time-based scheduling may be required
for meeting the demands of on-time execution. Implementation of on-time
systems on distributed platforms and with fault tolerant capabilities
poses many challenges. In this tutorial the primary approaches to the
design and implementation will be presented with some examples from the
Maruti system which is being implemented at the University of Maryland.

******************************************************************************
Monday, May 15, 3:30 - 5:00 (Advanced Track)

Advanced Topics in ATM Networks

Anujan Varma
University of California, Santa Cruz

This tutorial covers a number of advanced topics of current interest in
broadband ATM networks. Recent research results on these topics are
presented and, whenever possible, their applications in commercial
products are discussed. Topics include: introduction (review of ATM
concepts, discussion of key problems and challenges); traffic scheduling
algorithms for ATM switches (algorithm tradeoffs, FIFO scheduling and
its variants, fair queueing, virtual clock, jitter control algorithms);
congestion control in ATM networks (rate control schemes, cell discard
schemes, hop-by-hop flow control, experiences with TCP over ATM).

******************************************************************************
Tuesday, May 16, 8:30 - 12:00 (Introductory Track)

Performance Evaluation 102: Workload Characterization Techniques

Tommy Wagner
United States Military Academy

In order to model the performance of a computer system two models are
actually required. One is a model of the system itself. The second
model is a model of the workload that runs on the system. The model
of the workload is typically used as input to the system model. For
example, the demands that the workload places on the various servers
in a system would be used as input to a queueing model of that system.
Benchmarking is another example of workload modeling. The process of
workload characterization is the process of building those workload
models. This tutorial will survey the field of workload
characterization focusing on some of the most widely accepted methods
for building workload models. Some of the tools available will be
discussed. As parallel system models have been developed, traditional
workload modeling methods have fallen short. New techniques are
required. The basic issues involved in parallel workload modeling will
be introduced and some of the proposed approaches will be discussed.

******************************************************************************
Tuesday, May 16, 8:30 - 10:00 (Intermediate Track)

Capacity Planning in Client/Server Environments

Daniel A. Menasce'
George Mason University

Through a motivating example of downsizing a mission critical
application to a client/server environment, the fundamental steps
involved in capacity planning for client/server environments are
presented. Performance modeling via analytic models is at the heart
of the methodology discussed. The tutorial shows how Mean Value
Analysis (MVA) and hierarchical modeling can be used to predict the
performance of client/server systems. A review of multiple class
MVA for closed queueing networks (QNs) with load dependent servers
is included in the tutorial. Several case studies will be used to
illustrate how local area networks (LANs), network interconnection
devices, and servers should be modeled when conducting capacity
planning for client/server systems.

******************************************************************************
Tuesday, May 16, 8:30 - 10:00 (Advanced Track)

Quality of Service (QoS) in High Speed Networks

Don Towsley
University of Massachusetts

Increases in bandwidths and processing capabilities of future packet
switched networks will give rise to a dramatic increase in the types
of applications using them. Many of these applications (e.g., audio,
video) will require guaranteed quality of service (QoS) such as a
bounded maximum end-to-end packet delay and probability of packet loss.
This poses exciting challenges to network designers. In this tutorial
we discuss the QoS requirements of different applications and survey
recent developments in the areas of call admission, link scheduling,
and the roles that they play in providing guaranteed QoS. We will
also discuss the interaction between the provision of QoS, call
routing, and traffic monitoring and policing. Last, if time permits,
we will describe recent developments regarding the movement of ideas
regarding QoS from networks to video servers, multimedia systems, and
continuous media databases.

******************************************************************************
Tuesday, May 16, 10:30 - 12:00 (Intermediate Track)

Simulation Models of Computer Systems

Herb Schwetman
Mesquite Software, Inc.

Simulation models of computer systems are important tools used by
analysts as they seek to identify and remedy system performance
problems. Moreover, models are often the only tool which can be used
in the system design and configuration stages, when no real systems
yet exists. This tutorial will introduce discrete-event simulation,
with an emphasis on process-oriented, discrete-event simulation.
Numerous examples based on computer systems (both hardware and
software components) will be presented. A brief review of some of
the commercially available tools will follow. The session will end
with a detailed example featuring a model of transaction processing
in a client-server system.

******************************************************************************
Tuesday, May 16, 10:30 - 12:00 (Advanced Track)

Efficient Solution of Large Steady-State Markov Chains

Graham Horton and Scott Leutenegger
Erlangen University University of Denver

Many performance modeling studies and tools require the steady-state
solution of Markov chains. Several methods are available for the
steady-state analysis of Markov chains which range considerably in
their algorithmic complexity and in their computational characteristics.
In particular, one often finds a variation in computation time of
integer factors, sometimes even orders of magnitude, between the
different schemes. This tutorial covers the most important solution
algorithms, both established schemes and more modern approaches. The
tutorial will be illustrated by practical examples and will cover the
following material: introduction (Markov chains, steady-state analysis);
direct methods (Gaussian elimination, LU decomposition, GTH algorithm);
basic iterative schemes (Gauss-Seidel, power, SOR methods); NCD Markov
chains and iterative aggregation/disaggregation schemes; the multi-level
algorithm; adaptive relaxation; parallelizability of the schemes;
computational comparisons of the methods using examples. The direct
methods will only receive cursory treatment since we focus on the
solution of large Markov chains, thus necessitating iterative methods.

******************************************************************************
Tuesday, May 16, 1:30 - 5:00 (Introductory Track)

Performance Evaluation 103: Basic Applications to Parallel Systems

Brian Carlson
Dakota State University

Parallel programs have properties which are quite different from their
sequential counterparts. Thus, basic performance modeling techniques
differ considerably with modeling and workload characterization
of sequential programs. This tutorial will examine basic issues in the
performance evaluation of parallel programs. Characterizations of
parallel programs include Amdahl's Law, average parallelism, speedup,
efficiency, parallelism profiles, execution signatures, and task graphs.
In many instances this information may be used to make scheduling
decisions. Often, obtaining the characterization information is
difficult and requires additional tools. One such tool that will be
described in detail is PICL. PICL provides performance information
to visualization systems such as ParaGraph, which will also be
discussed. Also, the characteristics of a parallel program may change
as it is executed. One algorithm for identifying these changes will
be described which manipulates data created by PICL. A series of case
studies will show how the algorithm works. A particular case study
taken from an Intel Paragon will show the implications of executing a
program with different levels of parallelism. ParaGraph will be used
to show the effect of parallelism for an application which was
instrumented using PICL. In the case study, consequences of Amdahl's
Law will be discussed.

******************************************************************************
Tuesday, May 16, 1:30 - 3:00 (Intermediate Track)

Evaluation and Design of Benchmark Suites

Jozo J. Dujmovic
San Francisco State University

Quantitative methods for evaluation and design of benchmark suites are
necessary, both for those who create benchmark suites and for those
interested in a proper interpretation of performance measurement results.
This tutorial includes a complete presentation of a recently developed
black-box approach to workload characterization, its application to the
evaluation and design of benchmark suites, and practical examples based
on the most popular industry standard benchmark suites. The benchmark
evaluation methodology is intended to help system analysts minimize the
cost of benchmarking by properly selecting an optimum subset of relevant
benchmark programs. The benchmark design methodology is intended to help
designers of industrial benchmark suites to quantitatively control the
features of their products. This tutorial will include the following
topics: (1) a survey of industry standard benchmark suites, (2) white-box
and black-box workload characterization methods, (3) benchmark program
space (a metric space where each point represents a benchmark program),
(4) evaluation of benchmark suites using the concepts of size, redundancy,
completeness, density, and granularity, (5) benchmark suite design using
the methods of optimum subsets and optimum weights, and (6) case studies
of benchmark suite evaluation, design, and evolution (SPEC benchmarks,
Livermore Fortran Kernels, and others).

******************************************************************************
Tuesday, May 16, 1:30 - 3:00 (Advanced Track)

Process Algebras and/or Petri Nets
for Integrated Performance Modelling

U. Herzog and M. Ribaudo
University of Erlangen-Nuernberg University of Torino

Stochastic Process Algebras (SPA) represent a novel approach to the
integration of qualitative and quantitative modelling, based on an
extension of the process algebra modelling methodology. This extension
has been recently demonstrated to have many interesting features. It
allows system descriptions that include quantitative information, which
may be used to predict the performance or dependability of the system,
in addition to qualitative information, which may be exploited when
reasoning about the functional behaviour of the system (e.g., finding
deadlocks or exhibiting equivalences between subcomponents). While
qualitative analysis is carried out using standard techniques for
process algebras, quantitative analysis is based on an underlying
continuous time Markov chain (CTMC). Generalized Stochastic Petri
Nets (GSPN) provide another, well known formalism that can also be
used to study in a single environment both qualitative and quantitative
behaviour of systems. This tutorial presents a comparison of the
GSPN and SPA formalisms in terms of the facilities that they provide
to the modeller considering both the definition and the analysis of
the performance model. Our goal is to provide a better understanding
of both formalisms, and to prepare a fertile ground for exchanging
ideas and techniques between the two. To illustrate similarities and
differences, we make the different issues more concrete by means of
an example taken from the field of communication systems. Topics
include: introduction and motivation; explanation of SPA approach
and features; brief revision of GSPN features; comparison of modelling
style in the two formalisms; explanation of qualitative and quantitative
analysis of both; detailed case study; summary and conclusions.

******************************************************************************
Tuesday, May 16, 3:30 - 5:00 (Intermediate Track)

An Introduction to Statistical Hypothesis Testing

Joseph L. Hellerstein
IBM Thomas J. Watson Research Center

This tutorial provides an introduction to statistical hypothesis testing,
a formal approach to using measurement data for deciding between
competing alternatives. A statistical test determines which of two
hypotheses (typically in the form of probability distributions) best
explains the measurement data. The H0 hypothesis represents the status
quo; the H1 hypothesis is the alternative that the experimenter hopes to
prove (e.g., the superiority of a scheduling algorithm). Test
construction involves specifying a test statistic (e.g., the sample
mean) and determining its critical region (values of the test statistic
for which H1 is accepted). Tests are evaluated based on the probability
of making a wrong decision, either false positives or false negatives.
Ideally, critical regions should be chosen so that both kinds of error
occur with low probability. However, in general, changing the critical
region so as to decrease one kind of error will increase the other.
Topics addressed in the tutorial include: significance levels, power
curves, the Neyman-Pearson Lemma for most powerful tests, and uniformly
most powerful tests.

******************************************************************************
Tuesday, May 16, 3:30 - 5:00 (Advanced Track)

Performance Modeling Issues in Financial Engineering

Randolph Nelson
OTA Limited Partnership

Financial Derivatives -- you might think that these were four letter
words judging from all of the recent press. Every day it seems one
reads about another firm being taken to the cleaners with fancy exotic
financial instruments. What are these instruments and how do you
determine the price that you should pay for them? These are a couple
of the questions that we address in this tutorial. Besides describing
some equity instruments, we show that pricing them is an application
of applied probability and uses tools familiar to the performance
modeler -- probability, stochastic processes, approximation and
simulation. To illustrate the similarities between performance
modeling and financial engineering we will derive an expression for
the price of a derivative and show how this is a first step in actual
trading. Along the way we show how arbitraguers use derivatives to
earn a risk-free profit from temporary market mispricings without
risking any capital.

******************************************************************************

==============================================================================
TECHNICAL PROGRAM, MAY 17-19, 1995

Wednesday May 17, 95
--------------------

8:00-9:00 Registration

9:00-9:15 Opening Remarks

9:15-10:15 MULTIPROCESSORS:
Mary Vernon, University of Wisconsin (Chair)

Memory System Performance of UNIX on CC-NUMA Multiprocessors
John Chapin, Stephen A. Herrod, Mendel Rosenblum, Anoop Gupta
Stanford University

Talisman: Fast and Accurate Multicomputer Simulation
Robert Bedichek
Massachusetts Institute of Technology

10:15-10:40 Break

10:40-12:00 MULTIMEDIA SERVERS:
Rick Bunt, University of Saskatchewan (Chair)

Reducing I/O Demand in Video-On-Demand Storage Servers
L. Golubchik, University of California - Los Angeles
J.C.S. Lui, Chineese University of Hong Kong
R.R. Muntz University of California - Los Angeles

On Configuring a Single Disk Continuous Media Server
Shahram Ghandeharizadeh, Seon Ho Kim, Cyrus Shahabi
University of Southern California

A Traffic Model for MPEG-Coded VBR Streams
Marwan Krunz, Herman Hughes
Michigan State University

12:00 - 1:00 Lunch

1:00 - 2:00 POSTER SESSION (See below for titles)

2:00 - 3:00 HOT TOPICS

I NETWORK TRAFFIC MEASUREMENT AND MODELLING:
Carey Williamson, University of Saskatchewan (Chair)

Panel
Walter Willinger, Bellcore
Vern Paxson, Lawrence Berkeley Laboratory
Benjamin Melamed, NEC
Peter Danzig, University of Southern California

II G-NETWORKS - RECENT ADVANCES:
Erol Gelenbe, Duke University (Chair)

Panel
Peter Harrison, Imperial College
Edwige Pitel, Imperial College
Onno Boxma, CWI, Amsterdam
Jean-Michel Fourneau, University of Versailles

III FREE PERFORMANCE AND FAULT TOLERANCE:
Srini Tridandapani, University of California, Davis (Chair)

Panel
Anton T. Dahbura, Motorola
Arun K. Somani, University of Washington
Charles Martel, University of California, Davis
John Mathews, University of California, Davis

IV DATA INTERPRETATION AND EXPERIMENT PLANNING IN PERFORMANCE TOOLS:
Allen D. Malony, University of Oregon (Chair)

Panel
B. Robert Helm, University of Oregon
Jeffery K. Hollingsworth, University of Maryland
Barton P. Miller, University of Wisconsin
Karsten Schwan, Georgia Institute of Technology

3:00 - 3:30 Break

3:30-4:30 FAULT TOLERANCE:
Kishor Trivedi, Duke University (Chair)

A Case for Two-Level Distributed Recovery Schemes
Nitin H. Vaidya
Texas A&M University

An Analysis of Decay-Usage Scheduling in Multiprocessors
D.H.J.Epema
Delft University of Technology

4:30 - 5:30 VIDEO MODELLING AND NETWORK CONTROL:
Ulrich Herzog, Universitaet Erlangen-Nuernberg (Chair)

Fundamental Results on the Performance of ATM Multiplexers with Applications to
Video Teleconferencing
Anwar Elwalid, AT&T Bell Laboratories
Daniel Heyman, Bell Communications Research
T. V. Lakshman, Bell Communications Research
Debasis Mitra, AT&T Bell Laboratories
Alan Weiss, AT&T Bell Laboratories

Fundamental Limits and Tradeoffs of Providing Deterministic Guarantees to VBR
Video Traffic
Edward W. Knightly, University of California - Berkeley
Dallas E. Wrege, University of Virginia
Jorg Liebeherr, University of Virginia
Hui Zhang, Carnegie Mellon University

5:30 - 6:30 SIGMETRICS TOWN MEETING

Thursday May 18, 1995
---------------------

8:30 - 10:00 QUEUEING APPLICATIONS:
Phil Heidelberger, IBM (Chair)

Exponential Bounds for the Waiting Time Distribution in Markovian Queues, with
Applications to TES/GI/1 Systems
Youjian Fang, Michael Devetsikiotis, Ioannis Lambadaris, A. Roger Kaye
Carleton University

Optimal Probabilistic Allocation of Customer Types to Servers
S.C.Borst
CWI, Amsterdam

Z-Iteration: A Simple Method for Throughput Estimation in Time-Dependent
Multi-Class Systems
Ibrahim Matta, A. Udaya Shankar
University of Maryland - College Park

10:00 - 10:30 Break

10:30 - 12:00 DISK SYSTEMS:
Dan Reed, University of Illinois (Chair)

Striping in a RAID Level 5 Disk Array
Peter M. Chen, University of Michigan
Edward K. Lee, DEC SRC

On-Line Extraction of SCSI Disk Drive Parameters
Bruce L. Worthington, University of Michigan
Gregory R. Ganger, University of Michigan
Yale N. Patt, University of Michigan
John Wilkes, Hewlett-Packard Laboratories

DASD Dancing: A Disk Load Balancing Optimization Scheme for Video-on-Demand
Computer Systems
Joel Wolf, Hadas Shachnai, Philip Yu
IBM T.J. Watson Research Center

12:00 - 1:00 Lunch

1:00 - 2:30 CACHE PERFORMANCE:
Ed Lazowska, University of Washington (Chair)

An Analytic Study of Dynamic Hardware and Software Cache Coherence Strategies
Harjinder S. Sandhu, Kenneth C. Sevcik
University of Toronto

SM-prof: A Tool to Visualize and Find Cache Coherence Performance Bottlenecks
in Multiprocessor Programs
Mats Brorsson
Lund University

A Study of Integrated Prefetching and Caching Strategies
Pei Cao, Princeton University
Edward W. Felten, Princeton University
Anna Karlin, University of Washington
Kai Li, Princeton University

2:30 - 3:00 Break

3:00 - 4:30 MEMORY AND PARALLEL SYSTEMS:
Daniel Menasce, George Mason University (Chair)

On Characterizing Bandwidth Requirements of Parallel Applications
Anand Sivasubramaniam, Aman Singla, Umakishore Ramachandran, H. Venkateswaran
Georgia Institute of Technology

Scheduling Memory Constrained Jobs on Distributed Memory Parallel Computers
Cathy McCann, Tera Computer Corporation
John Zahorjan, University of Washington

Active Memory: A New Abstraction for Memory System Simulation
Alvin R. Lebeck, David A. Wood
University of Wisconsin - Madison

4:30 - 5:30 WORK IN PROGRESS (Contact Satish Tripathi during the conference)
Conference participants share this time slot to present recent results.

CONFERENCE BANQUET

Friday May 19, 1995
-------------------

8:30 - 10:00 STOCHASTIC SYSTEMS:
Onno Boxmaa, CWI (Chair)

Calculating Transient Distributions of Cumulative Reward
Edmundo De Souza e Silva, Federal University of Rio de Janeiro, NCE
H. Richard Gail, IBM T.J. Watson Research Center
Reinaldo Vallejos Campos, Pontificia Universidade Catolica do Rio de Janeiro

Regenerative Randomization: Theory and Application Examples
Juan A.Carrasco, Angel Calderon
Universitat Politecnica de Catalunya

Computational Techniques for Accurate Processor Evaluation of Multirate,
Multihop Communication Networks
Albert G. Greenberg, R. Srikant
AT&T Bell Laboratories

10:00 - 10:30 Break

10:30 - 11:30 HOT TOPICS

I INTERNET - TCP - ATM:
Teun Ott, Bellcore (Chair)

Panel
Robert Braden, Information Sciences Institute
T.V. Lakshman, Bellcore
U. Madhow, University of Illinois
C. Brazdziunas, Bellcore
A. Broscius, Bellcore
A. Neidhardt, Bellcore

II NON-MARKOVIAN PETRI NETS:
Kishor Trivedi, Duke University (Chair)

Panel
A Bobbio, University of Brecia
M. Telek, Technical University of Budapest
R. German, Technical University of Berlin
G. Ciardo, College of William and Mary
Antonio Puliafito, University of Catania

III PERFORMANCE IMPACT OF SELF-SIMILARITY IN TRAFFIC:
Ashok Erramilli, Bellcore (Chair)

Panel
Walter Willinger, Bellcore
T.V. Lakshman, Bellcore
Dan Heyman, Bellcore
Amarnath Mukherjee, Georgia Tech
San-qi Li, University of Texas at Austin

11:30 - 12:30 POSTERS

12:30 - 1:30 Lunch

1:30 - 3:00 MEASUREMENT AND ANALYSIS:
Guenter Haring, University of Vienna (Chair)

The Interaction of Parallel and Sequential Workloads on a Network
of Workstations
Remzi H. Arpaci
Andrea C. Dusseau
Amin M. Vahdat
Lok T. Liu
Thomas E. Anderson
David A. Patterson
University of California - Berkeley

Disk-Tape Joins: Synchronizing Disk and Tape Access
Jussi Myllymaki, Miron Livny
University of Wisconsin - Madison

An Inter-Reference Gap Model for Temporal Locality in Program Behavior
Vidyadhar Phalke, B. Gopinath
Rutgers University

END OF CONFERENCE

==============================================================================
BOOK DISPLAY

There will be a display of recent and classic books in Performance
Evaluation, ranging from beginning textbooks to research monographs.

==============================================================================
TITLES of POSTER PAPERS (See Wednesday May 17 1-2, Friday May 19 11:30-12:30):

Batch Class Process Scheduler for Unix SVR4
J.Braams, Unisys Netherlands

State Space Reductions Using Stochastic Well-Formed Net Simplification:
An Application to Random Polling Systems
S.Donatelli, G.Franceschinis
University of Torino

Approximate Response Time Distribution in Fork and Join Systems
S.Balsamo, I.Mura
University of Pisa

A Semi-Empirical Approach to Scalability Study
Xiaodong Zhang, Zhichen Xu
University of Texas at San Antonio

PEDCAD: A Framework for Performance Evaluation of Object Database Applications
E. Hughes, M. Winslett
University of Illinois

Scheduling for Cache Affinity in Parallelized Communication Protocols
James D. Salehi, James F. Kurose, Don F. Towsley
University of Massachusetts

Modeling and Analysis of Multi Channel Asymmetric Packet Switch
Models in a Bursty and Non-Uniform Traffic Environment
Amit K. Chaterjee, Vijay K. Konangi
Cleveland State University

Timepatch: A Novel Technique for the Parallel Simulation of
Multiprocessor Caches
Gautam Shah, Umakishore Ramachandran, Richard Fujimoto
Georgia Institute of Technology

Future Applicability of Bus-Based Shared Memory Multiprocessors
Derek Eager C.R.M. Sundaram,
University of Saskatchewan

Modeling A Fibre Channel Switch with Stochastic Petri Nets
Gianfranco Ciardo, College of William and Mary
Ludmila Cherkasova, Valim Kotov, Thomas Rokicki
Hewlett-Packard Labs

A Prefetching Prototype for the Parallel File System on the
Intel Paragon and Its Performance Evaluation
Meenakshi Arunachalam, Alok Choudhary, Syracuse University
Brad Rullman, Intel

Very Computer

SIGMETRICS/PERFORMANCE '95 PRE-REGISTRATION

SIGMETRICS/PERFORMANCE '95 PRE-REGISTRATION