Efficient Interoperability and Execution of Logistics Modules in the Integrated Consumable Item Supp

1
Efficient Interoperability and Execution of
Logistics Modules in the Integrated Consumable
Item Support (ICIS) Model02F-SIW-101

• Gary Kratkiewicz Renu Kurien Bostwick BBN
  Technologies

Geoffrey Knauth BAE Systems
2
IntroductionGary KratkiewiczScientistBBN
Technologies
3
Purpose

• Learn about
• Converting conventional computer programs to
  Web-based ones
• A Java-based system for interoperation and
  execution of legacy logistics modules

4
Whats Special About ICIS
ICIS has saved the US military over 100 million
by computing reduced needs for stocking food and
fuels ICISs acceptance is due in large part to
its distributed desktop user interface
incorporating the Java Plug-In. ICIS analyses
involve gigabytes of data, which we now better
manage through a Java-based server-side module
for interoperation and execution of its logistics
modules
5
Presentation Agenda

• Overview of ICIS (Gary)
• Logistics Modules (Geoffrey)
• Data Set Manager (Renu)
• Wrap-Up (Gary)
• Questions (All)

6
Authors Qualifications

• The three authors
• Are software developers with extensive
  development experience in Java and other
  languages
• Have each worked on the ICIS project for several
  years each has led development of a significant
  ICIS module
• Have given seminars on aspects ICIS and the
  technology that underlies it

7
Overview of ICISGary KratkiewiczScientistBBN
Technologies
8
The Need for ICIS

• Defense Logistics Agency (DLA)
• Originally supplied low-cost, non-critical items
• Responsibilities greatly increased in 1990s
• Ensure personnel equipment readiness in
  humanitarian wartime military operations
• Issue Lacked visibility into potential logistics
  problems during future military operations
• When will I run out of stock?
• What should I buy today for tomorrows needs?
• Do I have adequate sources transportation?

9
What ICIS Does

• Models logistics aspects of simultaneous or
  overlapping future military operations
• Combat and humanitarian operations
• Accounts for competing normal activities
• Ties together a variety of data
• Plan from military commands
• Logistics data from DLA
• Equipment personnel data from the services
• Identifies potential logistics problems

10
Steps in an ICIS Analysis

• User, via GUI
• Lays out set of planned military operations
• Sets parameters
• Logistics Modules
• Compute demand
• Time-phased schedule and intensity of consumers
  (tanks, people, aircraft, ships, etc.)
• Failure and consumption rates of parts/items
• Source demands and identify problems
• Inventory and backorders
• Ordering, production, shipping lead times
• GUI displays results

11
The Core of the Problem

• The computations are simple...
• data is the issue
• Obtaining data on a timely basis from many
  military organizations
• Managing the gigabytes of input data, and the
  gigabyte of data per analysis
• Presenting the results data
• Java helps with the last two issues

12
Standalone ICIS

• Started October 1994 using rapid prototyping
  approach
• High-level requirements, but no specific ones
• Many iterations during the first few years
• Developed in Tcl/Tk on UNIX workstation, with
  computations soon moved to C
• Real use in June 1996 when installed on
  militarys secure version of Internet
• Emphasis on analysis data instead of GUI

13
Limitations of Standalone ICIS

• Remote performance unacceptable -- required
  installation of software and data on local
  workstation
• Such a system often not available
• Installation requires detailed configuration
• To meet accessibility issues, simple Web-based
  utilities developed
• Not well integrated with main system
• Data and computation issues

14
Moving to the Web

• Netscape Navigator 3 available in August 1996,
  with usable Java and JavaScript
• Shifting priorities limited Web development
• Parts of GUI moved to Web from 1996 to 1998
• Eventually, Tcl/Tk GUI dropped computations now
  launched via Unix command line, between set-up
  and results viewing on Web
• Full Web-based release in Feb 1999
• Achieved DII/COE Level 7 compliance in 1999
• Treated Web version no different than non-Web
  version

15
Redesign for Ease-of-Use and Efficiency

• Limitations of first web version
• Heterogeneous implementation
• Data and computation issues never addressed
• Attacked both the client side (GUI) and server
  side (data and computations)
• Reimplemented key portions of GUI using JFC/Swing
  via the Java Plug-In
• Used Java-based modules to implement a data
  management system to prevent duplication of data
  and computations

16
ICIS Architecture
Server
Client
ICIS GUI Edit Data Define Scenario Launch
Analysis View Results
Raw Data From Services Commands DLA
User
User- Provided Data
DSM GUI Add User Data
ICIS Support
17
GUI Redesign Using Java 2

• Targeted redesign at inadequate areas
• Advantages with moving to Java 2
• JFC/Swing components better GUI
• Consistent Java platform
• Not all users using standard desktop
• Code reuse
• Interactive GUI, not a bunch of pages
• Native look and feel
• Upgrade system without touching the desktop by
  updating the applets
• Disadvantage
• Requires users to install the Java Plug-In

18
Example Wizard -- Original
19
Example Wizard -- Redesigned
20
Example Editor
21
Another Example Editor
22
Logistics Modules Geoffrey KnauthSystems
AnalystBAE Systems
23
List of Modules

• B8/TPFDD Processing
• Bulk Fuels Demand Sourcing
• Subsistence Demand Sourcing
• Critical Repair Parts Demand
• Depot Repair Parts Demand
• Clothing Demand
• Construction and Barrier Demand
• Defensive Chemical Gear Demand
• Packaged Petroleum Parts Demand
• Peacetime Demand
• Repair/Replacement Parts Sourcing
• Repair/Replacement Parts Problem Items
  Computation Engine

24
Example Module Bulk Fuels

• Began as large C program Java used later
• Dozens of input files
• Primary modules are Requirements and Sourcing
• Thousands of output files
• Mostly web pages
• Some CSV files and logs
• Designed for deliberate planning

25
Moving Fuels Module to the Web

• Already web-based, had 50,000 links
• Did not yet have dynamic queries
• Typical user visited 20 to 200 links
• Users needed crisis action planning tool
• Initially added CGIs and JavaScript
• Now Java2 fulfills most of their wishes, and
  makes developers lives easier too
• Cleaner software, excellent kits
• Fewer platform issues

26
Fuels Module Redesign

• Write analyses as data, not web pages
• Focus on actions repeated by users
• Use CGIs and servlets for
• Modularity
• Flexible query capabilities
• Quick turnaround
• Access control
• Quality control

27
Conclusions

• Use server-side Java and modularize early for
  greatest adaptability
• Self-test with known and random data
• Monitor usage patterns
• Listen to user interface experts
• Test client-side software on all platforms
• Know whats in your JAR cache
• Use Javas strengths to impress your clients

28
Data Set ManagerRenu Kurien BostwickScientistB
BN Technologies
29
ICIS Server-Side Architecture
Server
Client
Data
Raw Data From Services Commands DLA
ICIS GUI Edit Data Define Scenario Launch
Analysis View Results
User
Data Set Manager
Preprocessing Modules
Processing Analysis Modules
User- Provided Data
DSM GUI Add User Data
ICIS Support
30
Processing Analysis Modules Legacy Code
31
Magnitude of the Problem isICIS Data

• Large amounts of data in files (800-1000 input
  files).
• 200 different formats
• Data files received from different agencies.
• Data files regularly updated.
• Data files moved around to free up disk-space.

32
Problems Solved

• Data Refresh Problem
• What-Ifs Problem
• Multi-Machine Problem

Data Set Manager Features

• Data module version traceability
• File computation dependency info stored
• Description-based filesystem instead of
  directory-based filesystem.

33
Description-Based Filesystem
Search Engine
File Predicate
Feature data tracing swapping
Feature machine-independent file
characterization
34
File Specs File Objects

Module

Module
Module


35
Module SpecsDefine a Computation
Module Spec

Module Version
Parameter Spec
Feature maintain module data dependency
36
Module Invocations Computations
Module Invocation
File Binding
Module Spec

Module Version
Parameter Spec
37
Computational Trees
Module Spec



File Binding
38
Computations are Stored
Feature computation re-use
Feature dependencies not machine-dependent
Search Engine
Module Spec
39
DSM Implementation

• Oracle Database (tables for File Objects, File
  Specs, Module Specs, Module Invocations, etc.)
  with JDBC access
• Jscheme for Module Spec definition language
• DSM is a server ICIS modules communicate with
  it via sockets using Scheme notation

40
DSM Conclusions

• ICIS DSM solves Data Refresh Problem, What-Ifs
  Problem, Multi-Machine Problem
• The use of Jscheme has allowed the seamless
  intertwining of Java classes and an
  application-specific description language.

ICIS uses Java with Jscheme Oracle on the
server to solve complex data management problems.
41
Wrap-UpGary KratkiewiczScientistBBN
Technologies
42
Successes

• One of the first operational Web-based systems in
  the US Department of Defense
• Saved over 100 million by computing reduced
  needs for stocking fuels and food, a tremendous
  return on investment
• A major military command saved 12.7 million over
  two years in reduced inventory requirements
• The Army saved 75 million by decreasing
  requirements for prepositioned subsistence
  inventory

43
Successes (Continued)

• A day-long analysis replaces calculations that
  previously took months to perform
• What-if capability leads to better decisions
• Role expanded to credible crisis action tool
• Used to assess readiness or compute operational
  requirements in military operations such as the
  following
• Hurricane Mitch relief
• DESERT FOX
• SOUTHERN WATCH
• Kosovo campaign

44
Successes (Continued)

• In 1999, identified in Joint Strategic Capability
  Plan as viable tool
• In 2000, documented in Joint Bulk Petroleum
  Doctrine
• Compatibility with other military software tools
  has saved their users time and effort
• Has identified missing equipment in military
  databases

45
Key Points

• Move traditional applications to the Web to gain
  tremendous benefits
• Far wider accessibility
• Use on almost any platform
• Deploy Java 2 on the desktop (via the Java
  Plug-In) to gain even more benefits
• Improve many aspects of your Web GUI
• Upgrade without touching the desktop by updating
  the applets

46
Key Points Continued

• A Java and Scheme-based Data Set Manager can
• Eliminate recomputation of identical intermediate
  results and duplication of data storage between
  analyses
• Allow virtually any legacy module to be added to
  the system
• These techniques are applicable to a wide range
  of software systems

47
Efficient Interoperability and Execution of
Logistics Modules in the Integrated Consumable
Item Support (ICIS) Model02F-SIW-101

• Gary Kratkiewicz Renu Kurien Bostwick BBN
  Technologies

Geoffrey Knauth BAE Systems