home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!olivea!hal.com!darkstar.UCSC.EDU!golding
- From: mohr@cip.informatik.uni-erlangen.de (Bernd Mohr)
- Newsgroups: comp.doc.techreports
- Subject: TR-List: Measurement+Modeling Parallel Systems, University of Erlangen
- Message-ID: <156ipfINNfcd@darkstar.UCSC.EDU>
- Date: 29 Jul 92 09:23:45 GMT
- Organization: Student Pool, CSD, University of Erlangen, Germany
- Lines: 390
- Approved: compdoc-techreports@ftp.cse.ucsc.edu
- NNTP-Posting-Host: midgard.ucsc.edu
- Originator: golding@midgard
-
- *************************************************************************
- * UNIVERSITY OF ERLANGEN *
- * Institute for Mathematical Machines and Data Processing (IMMD7) *
- * Group for Measurement, Modeling and Evaluation *
- * of Parallel and Distributed Systems (MMB) *
- * D-8520 Erlangen *
- * GERMANY *
- * *
- * DOCUMENTS FTP ACCESS *
- *************************************************************************
-
- host: faui79.informatik.uni-erlangen.de [131.188.47.79]
- directory: ~ftp/pub/doc
-
- In case of any problem or question feel free to send a mail to:
-
- mohr@immd7.informatik.uni-erlangen.de
-
- This directory contains documentation files in PostScript
- format (compressed) about the ZM4, SIMPLE, and PEPP environment.
- The entries are described in the following format:
-
- Filename (number of pages)
- "Title"
- Author(s)
- Reference
- Abstract
-
- The email addresses of the authors:
-
- Peter Dauphin pdauphin@immd7.informatik.uni-erlangen.de
- Franz Soetz franz@immd7.informatik.uni-erlangen.de
- Richard Hofmann rhofmann@immd7.informatik.uni-erlangen.de
- Rainer Klar klar@immd7.informatik.uni-erlangen.de
- Bernd Mohr mohr@immd7.informatik.uni-erlangen.de
- Andreas Quick quick@immd7.informatik.uni-erlangen.de
- Markus Siegle siegle@immd7.informatik.uni-erlangen.de
-
- **************
- * ARTICLES *
- **************
-
- conpar90.si.ps.Z (12 pages)
- "Performance Evaluation of Parallel Programs
- in Parallel and Distributed Systems"
- B. Mohr
- In H. Burkhart, editor, CONPAR 90-VAPP IV,
- Joint International Conf. on Vector and Parallel Processing.
- Proceedings, pages 176-187, Zurich, Switzerland, September
- 1990. Springer, Berlin, LNCS 457.
-
- This paper deals with performance evaluation of parallel
- and distributed systems based on monitoring of concurrent
- interdependent activities. First a model is introduced
- for describing the dynamic behavior of computer systems
- in terms of events. Then, a distributed hardware/hybrid
- monitor system based on event driven monitoring and
- its tool environment SIMPLE are presented. We emphasize
- the tool environment as a prerequisite for successful
- performance evaluation. The tool environment for evaluating
- event traces, which integrates the data access interface
- TDL/POET and a set of evaluation tools for processing
- the data, makes evaluation independent of the monitor
- device(s) and the object system. It provides a problem
- oriented way of accessing event traces.
-
- conpar90.pp.ps.Z (10 pages)
- "A Method for Performance Prediction of Parallel Programs"
- F. Soetz
- In H. Burkhart, editor, CONPAR 90-VAPP IV,
- Joint International Conf. on Vector and Parallel Processing.
- Proceedings, pages 98-107, Zurich, Switzerland, September
- 1990. Springer, Berlin, LNCS 457.
-
- In general there are many possibilities to parallelize an
- algorithm and to distribute the tasks to the processors. In
- some cases the selection of a processor configuration depends
- on the given problem, too. To decide which implementation for
- any configuration may be the best, the programmer can use
- methods of performance evaluation and prediction. If the
- structure of the parallel program can be modeled by a seriesparallel
- graph we know a method to obtain the runtime of the total program.
- This paper deals with the more difficult problem of predicting
- the runtimes of non-seriesparallel structured programs. It is shown
- how to compute approximately the average runtime of a program which
- consists of tasks with deterministically and/or exponentially
- distributed runtime variables. The method is based on the well known
- transient state space analysis.
-
- dagstuhl.ps.Z (28 pages)
- "Integrating Monitoring and Modeling to a
- Performance Evaluation Methodology"
- R. Hofmann, R. Klar, N. Luttenberger, B. Mohr, A. Quick, and F. Soetz
- In T. Haerder, H. Wedekind, and G. Zimmermann,
- editors, Entwurf und Betrieb verteilter Systeme, pages
- 122-149. Springer-Verlag, Berlin, IFB 264, 1990.
-
- This paper presents a comprehensive methodology for
- monitoring and modeling parallel and distributed systems
- systematically. The integration of models, measurements,
- and evaluators to an efficient set of performance evaluation
- tools is described. Three typical tools are presented.
- One of them is the distributed hardware and hybrid
- monitor ZM4, another is the monitor independent and
- source related event trace interface POET/TDL. Both
- were developed at the Universitaet Erlangen-Nuernberg.
- As a modeling tool stochastic Petri-nets have been
- used. These tools have been used for analyzing the
- performance of multiprocessor and multicomputer systems.
- Here, they are applied in a case study for performance
- analysis and improvement of a communication subsystem
- prototype for B(roadband)-ISDN that was developed by
- IBM's European Networking Center. The measurement results
- give some interesting hints concerning the prototype's
- architecture which helped to improve the communication
- subsystem. Measuring the existing communication subsystem
- was accompanied by models for predicting the performance
- of modified ones.
-
- edinburgh.ps.Z (19 pages)
- "Model-driven Validation of Parallel Programs Based on Event Traces"
- P. Dauphin, M. Kienow, and A. Quick
- In Proc. of Conf. on Programming Environments for Parallel
- Computing, Edinburgh, April 1992.
-
- In order to program parallel and distributed systems
- efficiently, a systematic way of defining and understanding
- the complex behavior of process interactions in concurrent
- programs is needed. A proved method for understanding
- existing programs is model-driven monitoring which
- abstracts the program behavior to a sequence of events.
- These event traces are analyzed for debugging and tuning
- the program. However, model and program are not always
- consistent. The goal of this paper is using measured
- event traces to validate the consistency between model
- and program.
- We have investigated a solution to validate model-program
- consistency automatically: by building a functional
- model of the parallel program describing all properties
- for debugging and performance evaluation the definition
- of monitoring events is automated and carried out systematically.
- This integration of modeling and monitoring guarantees
- the same set of events in monitoring and modeling.
- It enables us to validate the program behavior in an automatic
- and systematic way by checking the event trace against the behavior
- represented in the model.
- In order to implement validation for different modeling
- methods only once we analyzed three different methods
- for a model-type-independent representation of parallel
- programs. The analysis showed that a model of a real-world
- application is too complicated to be described in a
- model-type-independent way. Therefore validation was
- implemented for two modeling tools: PEPP for modeling
- with graph models and GreatSPN for modeling with Petri
- nets. Both tools were extended to enable automatic
- program validation.
-
- edmcc2.si.ps.Z (10 pages)
- "SIMPLE: a Performance Evaluation Tool Environment
- for Parallel and Distributed Systems"
- B. Mohr
- In A. Bode, editor, Proc. of the 2nd European Distributed
- Memory Computing Conference, EDMCC2, pages 80-89, Munich,
- Germany, April 1991. Springer, Berlin, LNCS 487.
-
- This paper describes SIMPLE: a performance evaluation
- tool environment for parallel and distributed systems
- based on monitoring of concurrent interdependent activities.
- We emphasize the tool environment as a prerequisite
- for successful performance evaluation. All tools use
- the data access interface TDL/POET which can decode
- measured data of arbitrary structure, format and representation.
- This makes the evaluation independent of the monitor
- device(s) used and the system monitored. It also provides
- a problem-oriented way of accessing the data. Therefore
- it is very easy to adapt SIMPLE to any kind of measured
- data and to understand the evaluation results.
-
- edmcc2.tr.ps.Z (10 pages)
- "Monitor-Supported Analysis of a Communication
- System for Transputer-Networks"
- C.-W. Oehlrich, A. Quick, and P. Metzger
- In A. Bode, editor, Proc. of the 2nd European Distributed
- Memory Computer Conference, EDMCC2, pages 120-129, Munich,
- Germany, April 1991. Springer, Berlin, LNCS 487.
-
- Most parallel applications in a Transputer-network
- require a lot of communication between the processing
- nodes. For such applications the communication system
- TRACOS was developed to support data transfer between
- arbitrary Transputers in the network. To maximize the
- performance of the parallel system its dynamic internal
- behavior has to be analyzed. For this purpose event-driven
- monitoring is an appropriate technique. It reduces
- the dynamic behavior of the system to some important
- events, which are recorded by a monitor system and
- stored in event traces. In this paper the architecture
- of the communication system TRACOS and its analysis
- are presented. For the analysis a synthetic workload
- was instrumented and monitored with the distributed
- hardware monitor ZM4.
-
- ieee.ps.Z (26 pages)
- "ZM4/SIMPLE: a General Approach to Performance-Measurement
- and -Evaluation of Distributed Systems"
- P. Dauphin, R. Hofmann, R. Klar, B. Mohr, A. Quick, M. Siegle,
- and F. Soetz
- In T.L. Casavant and M. Singhal, editors, Advances in Distributed
- Computing: Concepts and Design. IEEE Computer Society Press, 1992.
-
- The performance of parallel and distributed systems
- is highly dependent on the degree of parallelism and
- the efficiency of their communication systems. Both,
- efficiently parallelizing big jobs and successfully
- designing high-speed communication systems, need insight
- into the dynamic behavior of at least two computers
- at a time. Getting insight is usually needed in debugging,
- here it is a means for improving performance.
- First, we present a comprehensive methodology for monitoring
- and modeling programs in parallel and distributed systems.
- In using event-driven monitoring and event-oriented
- models there is a common abstraction, the event, which
- enables us to integrate both approaches.
- A second part describes implementation concepts in
- hardware and software which render the methodology
- generally applicable and fruitful for practical performance
- evaluation problems. The hardware monitor ZM4 uses
- many distributed monitor agents and a global clock
- mechanism for achieving generality, and the evaluation
- environment SIMPLE uses new trace description and access
- principles which allow for accessing arbitrarily formatted
- traces, thus making standardized formats in the measured
- event traces superfluous.
-
- ieee.pads.ps.Z (23 pages)
- "Distributed Performance Monitoring:
- Methods, Tools, and Applications"
- R. Hofmann, R. Klar, B. Mohr, A. Quick, and M. Siegle
- Submitted to: IEEE Transactions on Parallel and
- Distributed Systems, 1992.
-
- A method for analyzing the functional behavior and
- the performance of programs in distributed systems
- is presented. We use hybrid monitoring, a technique
- which combines advantages of both software monitoring
- and hardware monitoring. It is shown that monitoring
- and modeling both rely on a common abstraction of a
- system's dynamic behavior, and therefore can be integrated
- to one comprehensive methodology. This methodology
- is supported by a set of tools for modeling, monitoring,
- and measurement evaluation. Here we describe a hardware
- monitor and a software package (ZM4/SIMPLE) which make
- our concepts available to programmers, assisting them
- in debugging and tuning of their code. A short survey
- of related monitor systems highlights the distinguishing
- features of our implementation. As an application of
- our monitoring and evaluation system, the analysis
- of a parallel ray tracing program running on the SUPRENUM
- multiprocessor is described. We show how monitoring
- helped to increase the performance of a parallel program
- dramatically.
-
- isca91.ps.Z (10 pages)
- "Performance Evaluation of a Communication System for
- Transputer-Networks Based on Monitored Event Traces"
- C.-W. Oehlrich and A. Quick
- ACM SIGARCH, 19(3):202-211, May 1991. Proc. of the
- 18th Int. Symp. on Computer Architecture, Toronto, May 27-30, 1991.
-
- Most parallel applications (e.g.image processing, multigrid
- algorithms) in a Transputer-network require a lot of
- communication between the processing nodes. For such
- applications the communication system TRACOS was developed
- to support data transfer between random Transputers
- in the network. To maximize the performance of the
- parallel system, its dynamic internal behavior has
- to be analyzed. For this purpose event-driven monitoring
- is an appropriate technique. It reduces the dynamic
- behavior of the system to sequences of events. They
- are recorded by a monitor system and stored as event
- traces. In this paper the communication system TRACOS
- and its performance evaluation based on monitored event
- traces are presented. First a synthetic workload was
- instrumented and monitored with the distributed hardware
- monitor ZM4. The results showed that the performance
- of TRACOS is poor for packets smaller than 4Kbyte.
- Therefore, TRACOS itself was instrumented and monitored
- to get insight into the interactions and interdependencies
- of all TRACOS processes. Based on the monitoring results,
- TRACOS could be improved which led to a performance
- increase of 25%.
-
- isca92.ps.Z (10 pages)
- "Monitoring Program Behaviour on SUPRENUM"
- M. Siegle and R. Hofmann
- ACM SIGARCH, 20(?):???, May 1992, Proc. of the 19th Int. Symp. on
- Computer Architecture, Queensland, May 19-21, 1992.
-
- It is often very difficult for programmers of parallel
- computers to understand how their parallel programs
- behave at execution time, because there is not enough
- insight into the interactions between concurrent activities
- in the parallel machine. Programmers do not only wish
- to obtain statistical information that can be supplied
- by profiling, for example. They need to have detailed
- knowledge about the functional behaviour of their programs.
- Considering performance aspects, they need timing information
- as well. Monitoring is a technique well suited to obtain
- information about both functional behaviour and timing.
- Global time information is essential for determining
- the chronological order of events on different nodes
- of a multiprocessor or of a distributed system, and for
- determining the duration of time intervals between events from
- different nodes. A major problem on multiprocessors is the absence
- of a global clock with high resolution. This problem can be overcome
- if a monitor system capable of supplying globally valid time
- stamps is used.
- In this paper, the behaviour and performance of a parallel
- program on the SUPRENUM multiprocessor is studied.
- The method used for gaining insight into the runtime
- behaviour of a parallel program is hybrid monitoring,
- a technique that combines advantages of both software
- monitoring and hardware monitoring. A novel interface
- makes it possible to measure program activities on
- SUPRENUM. The SUPRENUM system and the ZM4 hardware
- monitor are briefly described. The example program
- under study is a parallel ray tracer. We show that
- hybrid monitoring is an excellent method to provide
- programmers with valuable information for debugging
- and tuning of parallel programs.
-
- turin.ps.Z (15 pages)
- "Tools for a Model-driven Instrumentation for Monitoring"
- R. Klar, A. Quick, and F. Soetz
- In G. Balbo, editor, Proc. of the 5th Int. Conf. on Modelling
- Techniques and Tools for Computer Performance Evaluation,
- Torino, Italy, pages 165-180. Elsevier Science Publisher B.V., 1992.
-
- The performance of parallel and distributed systems
- depends significantly on how the programs' execution
- is dynamically organized. Therefore, performance evaluation
- and efficient programming of parallel and distributed
- systems need the analysis of the internal behavior
- of such systems. For this purpose event-driven monitoring
- is an appropriate technique. Event-driven monitoring
- reduces the dynamic internal behavior of the system
- to some important events, which are recorded in event
- traces. The instrumentation of the object program,
- i.e.the definition of events, depends on the respective
- intention of the analysis. In most cases the instrumentation
- will be done intuitively based on professional experience.
- In this paper a concept for a model-driven execution
- of event-driven monitoring is presented. It leads to
- tools for systematic and automatic instrumentation.
-
- ***********************
- * Reference manuals *
- ***********************
-
- simple.ps.Z (20 pages)
- SIMPLE User's Guide - Version 5.3
- SIMPLE Tutorial
- Technical Report no. ??/92
-
- tdl.ref.ps.Z (69 pages)
- SIMPLE User's Guide - Version 5.3
- Part A: TDL reference guide
- Technical Report no. 3a/92
-
- poet.ref.ps.Z (55 pages)
- SIMPLE User's Guide - Version 5.3
- Part B: POET reference manual
- Technical Report no. 3b/92
-
- tools.ps.Z (86 pages)
- SIMPLE User's Guide - Version 5.3
- Part C: Tools reference manual
- Technical Report no. 3c/92
-
- var-fil.ref.ps.Z (57 pages)
- SIMPLE User's Guide - Version 5.3
- Part D: FDL/VARUS reference guide
- Technical Report no. 3d/92
-
- pepp.ps.Z (54 pages)
- PEPP User's Guide
- Technical Report no. 5/92
- ===========================================================================
- Co-moderator: Richard Golding, Computer & Information Sciences, UC Santa Cruz
- compdoc-techreports-request@ftp.cse.ucsc.edu
-