Transcript Tutorial: Languages and Servers for Optimization Modeling

Productivity in the Enterprise through
OR-CI Synthesis and Integration
Workshop held in Washington, D.C.
August 30-31, 2004
Sponsored by the National Science Foundation
Organizers: Bob Fourer, Steve Wright, Jorge
Moore, Karthik Ramani
OR: Suvrajeet Sen
Shared CI: Sangtae Kim
OR as an Infrastructure
OR: Science of Decision making
Strengths
Integrates theory, algorithms, and software
Provides modeling and analysis tools
Underlies and facilitates productive activity in
design, manufacturing, services, supply-chain management
. . . serves many purposes not originally envisioned
Weaknesses
Lack of standards for interoperability of tools
Limited accessibility to modeling and analysis tools
. . . ad hoc, awkward interfaces
DMII Opportunities
Current Product Development Processes are extremely
iterative, communication intensive (data driven),
and linear.
Challenge areas to be addressed
Enterprise applications in such areas as
design optimization and configuration management
total supply network management
production planning over product lifecycles
simulation of decentralized services
Dynamic representations rather than static
. . . bypassed by Atkins report
Integrating OR within Cyberinfrastructure
Remedies for current weaknesses
Create an infrastructure to
enable research collaboration across
institutions, locations, time, and fields of
endeavor
Ensure that the data and software acquired at
great expense and effort are available for
future researchers
Replace incompatible software tools and
structures and take the lead in fostering
“coordinated” interoperability
Invest in maintenance and usability of
successful OR tools
. . . echoing dangers cited in Atkins report
Making OR a Cyberinfrastructure
Benefits of prompt action
Steer CI towards Meta Models, rather than complete
reliance on Meta Data
Focus time and talent on breaking new ground
rather than reproducing past efforts
Reduce isolation of the OR community
and among investigators within the community
Preserve-reuse valuable data (models and knowledge) for
future research
Combining top-down and bottom-up
approaches at multiple levels and scales
CYBER DOMAINS
DESIGN, MANUFACTURING, SUPPLY "NETWORKS"
LIFE CYCLE APPLICATIONS (CYBER-PHYSICAL INTERFACES)
(GRID-BASED COMMUNITIES)
OPERATIONS CI
ENTITIES: DATA-INFORMATION, KNOWLEDGE, ALGORITHMS,
SIMULATIONS, OPTIMIZATION-MODELS, CONFIGURATION
(linear, non-linear, combinatorial both determiniistic and stochastic)
OPERATIONS: LINKING, SEARCH, COMPARING, ANALYZING
OPERATING SYSTEM
PROCESSING
PHYSICAL LAYER
(COMMUNICATION INFRASTRUCTURE)
CYBER INFRASTRUCTURE PLATFORMS
Systems view of OR-CI
“High Level”
Architectures
System Level
User Interaction
& Interfaces
Applications
Community Resources
Functionality
Working level of interaction
standards with users and user
systems.
Interface to various computing
environments that enable access from
various current and future operating
platforms
Core tools for design, analysis,
manufacture and supply chain
coordination, etc. (Proprietary, open
source, and shared).
CADD, structural analysis programs, flow
visualization, simulation tools, enterprise
management, collaborative
communications, optimization etc.
Algorithms and analytical tools
available to the applications and user
levels.
“Low Level”
Architectures
Platform for community tool
development & management.
System Interoperation
Architecture
Examples
Low level operating software and
standards, security, and
communication protocols.
Data analysis tools, image processing,
statistical analysis functions, data
mining, search functions, language
translators, optimization languages.
Frameworks for interaction,
communications, and resource sharing
(DATA, INFORMATION, MODELS,
KNOWLEDGE).
CI-OR-EA and Engineering Design
Data/Knowledge/Models are the foundation of design
(repositories, libraries, catalogs)
CI and OR can facilitate:
 Access to remote and more current data/knowledge/models
 Organization of data, data mining and searching
methodologies
 More compute cycles => explore a much larger scenario
space, anticipate more exceptions and failure modes => more
robust designs
 More powerful, distributed algorithms for
• design under uncertainty
• design of flexible entities with many more degrees of
freedom
Supply-Network Management
The supply “chain”
Design and production
transportation and warehousing
marketing and delivery
Really a supply “network”
Part owned, part outsourced
Decision-making on wide range of time scales,
real-time control to short-term scheduling to long-term planning
Reconfigurable network
Robust optimization to deal with uncertainties
. . . a company must be able to easily use the network
Challenges in Supply-Network Management
Diverse tools required
Drawing on statistical, simulation, and optimization techniques
Communicating with each other,
with varied data sources,
with human analysts at different levels and locations
Difficult degree of integration required
Approached by some costly and specialized proprietary systems
Still out of the reach of most researchers and practitioners
. . . great potential for an operations cyberinfrastructure
CI-OR for Supply-Network Management
Standards for web services
Enable quick and reliable connections
between diverse analytical methods and data sources
Free time for experimentation
with new computational ideas and new software components
Accessibility of the CI
Speed integration of new research ideas into practice
Disseminate new supply-network ideas
to a broader variety of companies, especially relatively small ones
Promote use of the most challenging approaches, such as
optimization under uncertainty,
distributed simulation and optimization,
global optimization on noisy data,
optimization of simulations
Example: CI-OR in Product Lifecycle
Challenges of outsourcing
Steadily increasing demands for customization of products, but . . .
Further increases in complexity
Management of interfaces between suppliers
threatens to become expensive and inefficient
Dispersion of engineering and production
increases opportunities for breakdown in multi-tier supply networks
Hidden logistical and inventory costs and increased lead times
Consequences for design and development
Highly iterative
Communication-intensive
Reliant on suppliers from prototyping to production
Example: CI-OR in Product Customization
Consequences for the supply network
Significant time and cost to develop
a stable, reliable supply network for a product
Intensive coordination between different tiers
Rigid networks, unresponsive to dynamically changing markets
High inventory costs, borne mainly by
lower-tier suppliers already under pressure to cut costs
What CI-OR in enterprise applications can provide
Competitive advantages through
productivity improvements at all levels
More competitive based on
speed and responsiveness of the supply network
Solve large-scale distributed optimization and constraints
Handle large scale systems at multiple levels and scales
. . . not just for the biggest players …
CI-OR-EA and Enterprise Design and Services
Includes design of
 Physical entities (e.g. electrical grid, data networks)
 Virtual entities (alliances and markets)
Grid and network design:
 Design for robust (decentralized?) control, to allow for continuing operation
after disruptions (Big algorithmic challenges for OR)
 Placement of sensors, handling of sensor data are major issues
Market design (electricity markets, health information alliances):
 CI: Standards for information exchange
 OR: Use models/algorithms to design policies and pricing mechanisms to
facilitate efficient and fair operation
CI will enable the componentization of business infrastructures and
result in service oriented IT models
Pervasive connectivity between physical infrastructure and CI will
enable new service models
OR-Cyberinfrastructure: Example
Modeling System
Agent
Registry
Modeling Language
Local
Analyzer
Local
Centralized
Solver Interface
Function Evaluator
Distributed
Distributed
Solver
Function
Simulator
Distributed
Conclusions
The demands of real time and competitive decisions 
designed CI platform.
The CI-OR-EA computational engines and service specific
data repositories and libraries  increased DMS
productivity.
Innovation in design and manufacturing as well as R&D.
Capture commonalities and eliminate duplication 
increase quality and reliability.
OR – CI: resource location, network flow and assignment
problems.
Target applications in areas where technologies can be
gainfully employed.
CI and OR are essential partners to drive enterprise wide
productivity.