Title: The Wrong and the Right Way to Integrate Hardware, Software, and People
1The Wrong and the Right Way to Integrate
Hardware, Software, and People
- Arthur Pyster
- Senior Vice President and
- Director of Systems Engineering and Integration
- March 15, 2006
2Topics
- SAIC Profile
- Illustrative Examples of Wrong and Right Way to
Approach Building and Integrating Systems with
Significant Hardware, Software, and People
Requirements - Wrong
- Standard Terminal Automation Replacement System
(FAA) - Delphi Enterprise Resource Management System
(DOT) - Right
- Common Driver Trainer (Army)
- Mobile Inshore Undersea Warfare (Navy)
- Four Strategies and Their Rationale
- Conclusions
3SAIC Profile
- Homeland Security Co. of the Year Frost
Sullivan May 2005 - 1 IT Contractor at OSDINPUT January 2005
- 8 Largest DoD ContractorDoD January 2005
- 2 Federal IT Contractor INPUT April 2005
- 2 Top Systems Integrator Federal Computer Week
September 2005 - 3 Professional Services Firm FederalTimes.com
June 2005 - 3 Technology Contractor Government Executive
August 2005 - Leadership Top IT Outsourcing Vendor In North
AmericaMeta Group METAspectrum January 2004 - Leadership Desktop and Help Desk Outsourcing in
North AmericaGartner Magic Quadrant December
2004 - 3 Top Systems Integrator in North America
Gartner Dataquest August 2005 - Americas Most Admired Companies in Computer and
Data Services FORTUNE 2001-2005
O u r V i s i o n Be a leading systems and
solutions company, solving our customers most
important business and mission-critical problems
through innovative applications of technology and
domain knowledge From Science to Solutions
- O u r S u c c e s s e s
- More than 3 decades of continuous revenue growth
- 7.2 billion in annual revenues for FY05
- FORTUNE 500 company - 276
- 15.5 revenue CAGR over last 5 years
- Superb staff of qualified professionals
- More than 43,000 personnel worldwide
- 11,000 employees with advanced degrees
- 20,000 with security clearances
- Key positions on programs of national importance
- Including DoD transformation, border security,
intelligence analysis, cancer research and other
national priorities - More than 10,000 active contracts
- Leading provider of contracted RD services
- 3 billion in cash to support growth
4STRATEGIES for Effective Hardware,Software, and
People Integration
Generally, nothing flashy is required. Follow
modern systems engineering practices and you will
succeed.
Strategy 1. Keep users involved
throughout. Strategy 2. Embrace and manage the
business process reengineering that is inevitable
when large critical systems are
replaced. Strategy 3. Anticipate the next system
you will build and incrementally create a product
line architecture so that the next system (and
ones beyond it) can be more rapidly constructed
and fielded. Strategy 4. Build a little, test a
little, field a little. Do it often.
5First, the wrongs System developments
that used poor strategies
6STARS Overview
The primary automation to control an airplane
when it is beyond an airport and lower than
18,000 feet
1
CUSTOMER
A perfect example of how not to succeed
FACT 2,000,000 passengers are handled by TRACONS
every day
7Picture of Early ARTS, the STARS Predecessor
8Context in Which STARS was Created
- From early 1980s through early 1990s, FAA spent
4B to modernize air traffic control Advanced
Automation System - One of the most studied examples of disastrous
large-scale acquisition, systems engineering, and
software engineering - Big bang replacement planned for most legacy
systems controlling air traffic failed because
problem was much too complex, too poorly
understood, and too volatile - In 1996, new FAA acquisition management system
put in place separate from Federal Acquisition
Regulations - New strategy was set to modernize the National
Airspace System (NAS) incrementally with multiple
vendors, multiple acquisitions, driven by
enterprise architecture
9STARS Shone Dimly at First
- ARTS system, designed more than 20 years earlier,
was being used for low altitude air traffic
control across U.S. - In 1996, Raytheon was awarded 960M contract to
replace ARTS. - STARS was billed as a quick delivery system
because it would mostly be COTS with rollout
complete in 2004 or 2005. - Raytheon was to largely reuse system they had
deployed for Norway. - First rollout was to be in Boston in 1998.
- Air traffic controllers would not accept
capabilities and demanded extensive
customization. FAA management accepted demands
without understanding the implications. - STARS became major development effort, needing
hundreds of thousands of lines of new code. Cost
escalated to 1.4B. (Now up to 1.7B)
10User Interface Morass
- FAA requirements team, systems engineers, and
acquisition team were from different
organizations and didnt come together into an
effective integrated team. - Controllers had no effective way to articulate
their requirements. - FAA did not have a good way to elicit, negotiate,
or manage requirements. - There was little human factors expertise on the
FAA STARS team or, more broadly, within the FAA. - Controllers made demand after demand for features
without a good process for either side to
understand the consequences. - Engineers specified a controller interface that
was radically different from what controllers
wanted and were used to.
11User Interface Resolution
- Unlike ARTS, new display had no shortcut
switches and knobs. - Everything in STARS was to be controlled through
a trackball and keyboard it felt more like a
Windows PC interface. - Drop-down menus cluttered screen, blocking
controllers view of air traffic. - Controllers union went to Congress and the press
for changes. - After much finger-pointing between controllers
union and FAA, the FAA adopted new approach led
by their new Chief Scientist for Human Factors. - New process included prototyping, impact studies,
method to resolve conflicts between controllers
and program, . - FAA Administrator stayed personally involved in
regular STARS meetings for next several years.
12Picture of Early STARS
13STARS Backend Morass
- In addition to radically revising the frontend
user interface, the FAA faced numerous new
requirements from controllers and technicians for
new backend features around improved robustness,
maintenance, and performance. - Major rework of the backend processing system
would take years to complete. - Pressure mounted to get something in the field
quickly, but the STARS lifecycle was waterfall
development and incremental deployment.
14Backend Recovery
- New lifecycle was defined, enabling incremental
capability delivered to field. - FAA Administrator began to talk about build a
little, test a little, deploy a little. - STARS Early Display Configuration (EDC) defined
to have old ARTS backend with the new display
front-end. - STARS Full Scale-1 (FS-1) and Full Scale2 (FS-2)
defined incremental backend capability. - By 2001, STARS EDC was deployed at two low-volume
TRACONS El Paso and Syracuse. Recall original
roll out for STARS had first deployment at Boston
one of the busiest TRACONS in the country.
15Meanwhile
- Legacy system ARTS, that STARS was to replace,
kept advancing under a contract with Lockheed
Martin. - Common ARTS began to offer many of the advantages
that STARS was to offer, including large color
screens. - Critics begin to question whether the FAA should
abandon or cut back on STARS and field more
Common ARTS systems. - Eventually, FAA agreed to a hybrid with some
Common ARTS systems and some STARS systems. - Today, STARS deployment is proceeding, but wont
be completed for several more years. - Bostons STARS system was commissioned in May
2004.
16Picture of Common ARTS
17Delphi Enterprise Resource Planning
- Legacy system DAFIS built in 1970s as custom
application for Department of Transportation
(DOT) - Dozens of bolt-on applications developed over
the years to compensate for its shortfalls - Was recently replaced by new Delphi system that
was based on Oracle Federal Financials. - FAA is by far the dominant part of DOT, having
nearly all the complexity because it acquires
billions of dollars worth of systems annually and
operates a real-time safety-critical system worth
more than 25B. - DOT managed the acquisition of Delphi without a
good understanding of FAA complexities. The
primary acquisition leadership largely reflected
the other parts of DOT that managed grants,
investigated accidents, kept statistics, etc. The
gulf in perspectives was large.
18Business Process Reengineering Misunderstood
- DOT insisted FAA reengineer their business
processes in order to make effective use of
Delphi. - FAA leadership insisted that DOT didnt
understand the magnitude of the reengineering and
that DOT wasnt willing to fund it. - FAA line organizations had operated in a fairly
decentralized way using DAFIS with lots of
significant variation between organizational
elements in how financial and contractual
functions were performed. FAA line organizations
did not want to change their ways. - DOT was under great pressure from OMB to get
Delphi operational quickly. - FAA opted to minimize business process change in
the interests of speedy implementation and
because reaching consensus about new business
processes was very hard. - FAA didnt appreciate they couldnt avoid much of
the business process reengineering anyhow,
leading to gross underestimation of the amount of
time to convert to Delphi, including data
conversion and user training. Symptoms included
being behind in paying hundreds of millions of
dollars in bills.
19Next, the rights System developments that
followed smart strategies
20Common Driver Trainer Architecture
Advanced virtual simulators built from a common
architecture used to train Army personnel to
drive many types of vehicles
A model of how to succeed
FACT 272 casualties in operations Iraqi and
Enduring Freedom (through 12/3/05) were caused by
private automobile accidents and vehicle
crashesthe majority in non-hostile conditions.
21Modular Reconfiguration for Multiple Vehicles,
Scenarios, and Training Objectives
Product Line Architecture describes combined
software/hardware system
Combination of words and pictorials (Depiction of
the project in operation how it satisfies the
shortfall)
Product line includes instructor operator
station, training management system, automated
data collection and after action review
22Mobile Inshore Underwater Warfare
Provide surface and subsurface surveillance in
shore areas and provide supporting command and
control
FACT MIUW improve-ment and deployment has been
rapid since the attack on the USS Cole in 2000.
23Mobile Inshore UnderwaterWarfare Installation
24STRATEGIES for Effective Hardware,Software, and
People Integration
Generally, nothing flashy is required. Follow
modern systems engineering practices and you will
succeed.
Strategy 1. Keep users involved
throughout. Strategy 2. Embrace and manage the
business process reengineering that is inevitable
when large critical systems are
replaced. Strategy 3. Anticipate the next system
you will build and incrementally create a product
line architecture so that the next system (and
ones beyond it) can be more rapidly constructed
and fielded. Strategy 4. Build a little, test a
little, field a little. Do it often.
25Strategy 1. Keep Users Involved
- STARS failed to systematically involve
controllers from the beginning in a way that
built up their confidence and really analyzed
their needs, how they really operated their
equipment. This built up animosity and created
an atmosphere that took years to correct. - STARS failed to involve human factors experts
early in the lifecycle and had to recreate the
user interface later at great expense and time. - STARS failed to say no to users who wanted
features but didnt pay for them. - CDT recognized the gap between SEs who specify a
system and the actual users. SAIC specifiers are
usually twice the age of military system users.
Young soldiers have different mental models and
handle many tasks simultaneously. Use SEs with
characteristics of real users where possible. - MIUW recognized that their users are often
reservists, who dont get the same level of
training as regular forces. They needed to make
it possible to both use and maintain MIUW
equipment with minimal training. - MIUW used some of their own technicians in the
design e.g., technicians suggested that each
case has storage room for the cables that go with
that case in order to make setup and tear down
easier for the user. - MIUW observed that the resource loaded network is
a great communications device with the customer.
26Strategy 2. Embrace and Manage Change
- Delphi demonstrated that putting in place a major
new system will be disruptive to existing
business processes. - COTS systems bring constraints on business
processes that may clash with old business
processes. - Delphi demonstrated that senior leadership must
embrace the change in business process and then
manage it. - Delphi demonstrated that failure to embrace
change will lead to underestimating the costs of
training, the time for conversion, and the
disruption to current business operations.
27Strategy 3. Anticipate the Next System
- Product line architectures (PLAs) enable speedy
creation of individual systems. - There is little time or to create fully
generic PLAs outside real projects. Product
line is more a mindset than a toolset any
standard architectural tool can support this. - Begin with the order for a single product, but
anticipate a second product and design an
architecture for both. - Since first product is delivered, focus more
attention on ensuring PLA supports it. Support
for second product is more notional. Second
product proves architectural robustness. - Keep asking how to build next product while
building current one improve product line
architecture incrementally. - Keep line between HW and SW as fluid as possible.
28Strategy 4. Build quickly. Field often.
Impatience is a virtue. Dave Pratt
- STARS initially tried to get all requirements
correct and define a big bang delivery. They
backed into a spiral design with incremental
delivery. By backing into that strategy, the
program looked weak. Had they embraced it from
the beginning, they would have looked smart and
been smart. - CDT assumes that they wont get things right at
the beginning and fields trainers anticipating
rapid cycles of improvement. Trial and error and
rapid release is especially important for such
highly interactive systems as trainers. - MIUW fielded early prototypes to ensure they
really understood the requirements and that the
requirements were really right. They also
encountered the downside of this approach because
they had to keep the prototype in the field
longer than expected and it wasnt ruggedized. - MIUW used modeling, simulation, quality
assurance, and testing early and often to keep
quality high. QA and testing are especially
important for integration of components from
others.
29In Conclusion
Because
The big bang is dead. The waterfall is
dead. Ignored users kill systems. Impatience is a
virtue. Failure to anticipate leads to failure.
You must
- Keep your users close.
- Embrace change.
- Anticipate before you are asked.
- Build quickly, get more right than wrong, fix it,
repeat.