Future Generations of Problem-Solving Environments Jose C. Cunha - - PowerPoint PPT Presentation

▶

Apr 09, 2024 188 likes •524 views

' $ Future Generations of Problem-Solving Environments Jose C. Cunha Departament of Computer Science Faculty of Science and Technology New University of Lisbon, Portugal (jcc@di.fct.unl.pt) & % October 2000 ' $ Future

SLIDE 1 ' & $ %

Future Generations of Problem-Solving Environments

Jose’ C. Cunha

Departament of Computer Science Faculty of Science and Technology New University of Lisbon, Portugal (jcc@di.fct.unl.pt)

October 2000

SLIDE 2

Future Generations of Problem-Solving Envrionments 1

' & $ %

Index

1 st part:

Problem-Solving Environments

2 nd part:

Requirements for Future Generations of PSE

3 nd part:

Dimensions in PSE Development

4 nd part:

An Experience Towards Dynamic PSE

5 r d part:

Conclusions

SLIDE 3

Future Generations of Problem-Solving Envrionments 2

' & $ %

Problem–Solving Environments

Integrated environment supporting:

˘ an entire life cycle

development and execution steps to solve problems in a given application domain with easy access by an end–user

SLIDE 4

Future Generations of Problem-Solving Envrionments 3

' & $ %

Development Steps

Tools to help problem specification, design, analysis, verification,

evaluation: ˘ Rapid prototyping ˘ Dependent on a specific domain ˘ Expert assistance

SLIDE 5

Future Generations of Problem-Solving Envrionments 4

' & $ %

Execution Steps

To interact with ongoing experiments, by controlling and monitoring Activities performed on multiple heterogeneous components

(application–specific and generic tools): ˘ selection, evaluation and testing ˘ configuration, activation, interconnection ˘ monitoring, controlling

SLIDE 6

Future Generations of Problem-Solving Envrionments 5

' & $ %

Hetereogeneous Collection of Interconnected Components

Parallel problem solvers Expert assistance tools Tools for data processing, interpretation, visualization Tools for monitoring and computational steering Online access to large databases and scientific devices

SLIDE 7

Future Generations of Problem-Solving Envrionments 6

' & $ %

Requirements for Future Generations of PSE

Complex simulation models Large volume of input or generated data Difficult of their interpretation and classification

SLIDE 8

Future Generations of Problem-Solving Envrionments 7

' & $ %

End-user and Application Requirements

Higher Degrees of User Interaction Intelligence and Expert Assistance Tools Multidisciplinary Nature of the Applications

SLIDE 9

Future Generations of Problem-Solving Envrionments 8

' & $ % Higher Degrees of User Interaction

˘ User interfaces at distinct abstraction levels ˘ Increased flexibility in user and component interaction ˘ More advanced computational steering and visualization ˘ User driven and agent driven steering ˘ Distinct operation modes (offline/online data interpretation or visualization), dynamically selected by the user

SLIDE 10

Future Generations of Problem-Solving Envrionments 9

' & $ % Intelligence and Expert Assistance Tools

˘ Support for the development and the execution steps ˘ Advisoring/explaining tools to assist the user

During development time (correctness/performance) During execution time (impact of parameter modification upon system

behavior) ˘ Search for a balance between automated intelligent tools and an adequate level of user interaction

SLIDE 11

Future Generations of Problem-Solving Envrionments 10

' & $ % Multidisciplinary Nature of the Applications

˘ Support for interaction between distinct sub-models ˘ Support for distributed collaborative environments

SLIDE 12

Future Generations of Problem-Solving Envrionments 11

' & $ %

PSE System Requirements

Infrastructures for PSE Software Architectures Support for building PSEs Dynamic configuration and coordination isses

SLIDE 13

Future Generations of Problem-Solving Envrionments 12

' & $ % Infrastructures for PSE

˘ Low-level and middleware layers: towards meta-level distributed

perating systems and services

˘ Heterogeneity at the component level ˘ Operation at small and large scales ˘ Security issues ˘ Resource management and system configuration ˘ Cluster and metacomputing

SLIDE 14

Future Generations of Problem-Solving Envrionments 13

' & $ % Software Architectures

˘ To adapt the PSE and the tools according to the user’s interest ˘ Based on reuse of components and their dynamic modification ˘ Models for abstract specification of PSE ˘ Tools to reason about global system properties ˘ Tools to support transformation between software level

SLIDE 15

Future Generations of Problem-Solving Envrionments 14

' & $ % Support for building PSEs

˘ From manually assembled PSEs ˘ Towards automating their generation ˘ To handle their increased flexibility, complexity and size ˘ Meta environments for generating specific working PSEs

SLIDE 16

Future Generations of Problem-Solving Envrionments 15

' & $ % Dynamic configuration and coordination isses

˘ Dynamic component integration ˘ Modification of their interaction patterns ˘ Rely on the design of abstract interaction patterns ˘ Rely on dynamic reconfiguration of software architectures ˘ Raise new component and tool coordination issues ˘ Multiple users concurrently join ongoing experiments with distinct roles (observers, controllers) ˘ Provide consistency among views ˘ Provide answers to the distributed and dynamic nature of PSE components ˘ Provide answers to the need to dynamically adjust their interactions,

SLIDE 17

Future Generations of Problem-Solving Envrionments 16

' & $ %

depending on the user needs, the evolution of the experiments, and the system behavior

SLIDE 18

Future Generations of Problem-Solving Envrionments 17

' & $ %

Dimensions in PSE Development

SLIDE 19

Future Generations of Problem-Solving Envrionments 18

' & $ %

PSE

Application Components Coordination Software Architecture Monitoring and Control Resource Manag. Interconnection Infrastructures

T O O L S

Figure 1: Conceptual Layers

SLIDE 20

Future Generations of Problem-Solving Envrionments 19

' & $ % Coordination

˘ Represent and manage patterns of interaction among components ˘ Define cooperation and communication models ˘ Guarantees of consistency

Software Architecture

˘ High-level specification of components, their composition, their interactions, for a given problem ˘ Modeling and reasoning on the global structure and behavior ˘ Semantics of interactions through the component connectors ˘ Specification languages for:

Description of system structure and analysis of system behavior Incremental refinement and composition of architectures

SLIDE 21

Future Generations of Problem-Solving Envrionments 20

' & $ % Monitoring and Control

˘ Observation and control of distributed computations ˘ Distributed monitoring ˘ Computational steering ˘ Advanced visualization

Resource Management and Interconnection Services

˘ Configuration of parallel and distributed heterogeneous virtual machines ˘ Activation of component instances ˘ Mapping and load balancing ˘ Local scale and large scale operations ˘ Management of metacomputing resources

SLIDE 22

Future Generations of Problem-Solving Envrionments 21

' & $ %

˘ Component interconnection

Infrastructures

˘ Examples: Globus, Distributed Computational Labs, Generic PSEs

SLIDE 23

Future Generations of Problem-Solving Envrionments 22

' & $ %

Global Research Directions

Current status

˘ Build PSE for specific domains

Coperation with scientists / engineers Identification of user/application requirements Early and incremental development of prototypes Quick user feedback

˘ Make them evolve towards advanced PSE to ease development and execution of complex applications

SLIDE 24

Future Generations of Problem-Solving Envrionments 23

' & $ % Ongoing efforts

˘ Generic PSE to be tailored to specific problem domains ˘ Tools for the more/less automatic generation of application–specific PSE ˘ Integration of numeric, symbolic, multimedia, intelligent knowledge processing and discovery, database components

SLIDE 25

Future Generations of Problem-Solving Envrionments 24

' & $ %

Goals of Research at UNL

More flexible and dynamic PSE A framework to support parallel and distributed PSEs:

˘ Flexible and extensible tools for observation and control services ˘ Study the requirements for dynamic PSEs, their impact upon their software architecture, and the required coordination models

To use the framework to implement prototypes of specific PSEs and

evaluate application scenarios to assess dynamic configuration and coordination issues

SLIDE 26

Future Generations of Problem-Solving Envrionments 25

' & $ %

An Experiment Towards Dynamic PSEs at UNL

First based on a multidisciplinary Project

˘ Framework to support Parallel and Distributed PSE ˘ Tridimensional Optimal Layout of WasteWater Treatment Plants (WWTP)

SLIDE 27

Future Generations of Problem-Solving Envrionments 26

' & $ % Global Issues

˘ Integration of separate/distributed/heterogeneous components

distinct programming / computational models distinct / hybrid problem–solving strategies

˘ Parallel and distributed processing ˘ Interactive / adaptive control ˘ Easy access by the end–user in problem specification, development and execution control ˘ Dynamic reconfiguration ˘ Multiple cooperative users

SLIDE 28

Future Generations of Problem-Solving Envrionments 27

' & $ %

Integrated Environment

Global View. Several sub–models are coordinated by a central model, resulting in a completed computer aided design tool

Data exchange between sub–models and central model: central

model sends data (partial input) and gets results (partial output)

Interaction may use a subroutine style or communication between

independent processes

Parallelization is necessary for the optimization problem More information on this project and results in the paper and in:

http://www.cs.cf.ac.uk.euresco99/

SLIDE 29

Future Generations of Problem-Solving Envrionments 28

' & $ %

Experiments Towards Parallel Genetic Algorithms (GA) Environments

Data visualization: online evolution of the GA computation Interactive steering Adaptive control

SLIDE 30

Future Generations of Problem-Solving Envrionments 29

' & $ %

Experimentation: built several prototypes

for each separate component for their interconnection

SLIDE 31

Future Generations of Problem-Solving Envrionments 30

' & $ %

Conclusions: experiments on tools and mechanisms

To test and evaluate several parallel GA prototypes Use of a flexible monitoring and control architecture Use a distributed debugging tool for steering Use of a group based interconnection model

SLIDE 32

Future Generations of Problem-Solving Envrionments 31

' & $ %

Towards Dynamic PSEs

Issues

Dynamic component integration Distinct patterns of component interaction Increased flexibility in user and component interaction Component and tool coordination Multiple cooperative tools and users, sharing the state and controlling

an ongoing experiment Current tasks

How supporting dynamic reconfiguration can increase the flexibility of

a PSE, for an end-user

SLIDE 33

Future Generations of Problem-Solving Envrionments 32

' & $ % Design a collection of application level scenarios which involve

multiple tools and components of a PSE

Analyze how their dynamic reconfiguration can improve the

expressiveness of the life cycle for application development and

execution. Use specific case studies.

Model a diversity of interaction patterns among components and their

dynamic modification

Define a collection of coordination operations acting upon the