System of Systems: Power and Paradox

Joseph J. Simpson Dr. Cihan H. Dagli

Missouri University of Science and Technology Missouri University of Science and Technology
6400 32nd Ave NW #9 229 Eng. Mgt. & Systems Engineering Dept.
Seattle, WA 98107 Rolla, MO 65409-0370
jjs-sbw@eskimo.com dagli@mst.edu

Abstract – Systems concepts and artifacts provide the basis shown in Figure 1, Streams of Change [1]. While this is not
for enumerable sources of power and wealth in our modern an exhaustive list of change types, these five streams of
world. Culture, art and science all are based on established change are deemed sufficient to provide the necessary
systems of behavior, values and thought. The current context for the discussion of the design, development and
environment is densely populated with physical system production of technology-based, engineered systems.
artifacts that are used in every aspect of human life. The
ubiquitous nature of existing systems has generated a Science Technology Application Product Organization
strong interest in using an existing set of systems as the Stream Stream Stream Stream Stream

basis for a system of systems. Further interest in the

system-of-systems approach is stimulated by rapid
Obsolete Obsolete Obsolete Obsolete Obsolete
development, deployment and expansion of new and Science Technology Application Product Organization
existing systems. While successful system and system-of-
systems production provides the basis of great power, many
system development activities result in failure. The Current Current Current Current Current
Science Technology Application Product Organization
paradox is that while many individual systems work well as
a single system, they fail when incorporated as a
component of a system of systems. Successful system-of- New New New New New
systems characteristics and attributes are explored in this Science Technology Application Product Organization
paper.
t t
1
Keywords – System, System of Systems, System Success, Figure 1 Streams of Change
Paradox, System Failure, System Abstraction Frame,
The position of a system or system of systems can be
System Abstraction Stack.
generally located along the streams of change. This
position can be used to set the general system context.
1 Introduction
1.1 Abstraction Frames
All systems are developed in a dynamic context. Many
aspects of the system context are beyond the control and Using the sequential, flow-of-time context as stated in
influence of the system owner, architect, engineer and the introduction, a system evaluation framework has been
design engineer. The ability of a system, a system designed to support the discussion of systems and system-
component, or a system that is a member of a system of of-systems development. A system can be defined as a
systems to effectively sense and adapt to the dynamic relationship mapped over a set of objects [2]. During
system context is of great value. system design, system architects and engineers use their
Examples of large scale systems that must adapt to past experience and combined, technical knowledge to
dynamic context changes are reviewed in this paper. The design objects that, when properly integrated into a system
main objective of this review is the identification of system or system of systems, produce the customers desired
attributes and characteristics that facilitate the success of effects. The objects that make up a system must exist before
the system component, system and/or system of systems the system can be assembled. The abstraction frame
that are constructed and aligned in any given environment. concept was developed to encapsulate the flow of time in
The dynamic system context and environment will be one specific context [3]. There can be multiple, concurrent
represented by five general streams of change: the science abstraction frames active in any given system development.
stream, the technology stream, the application stream, the When objects and/or systems are instantiated in a single
product stream and the organizational stream. These are abstraction frame, then the time sequence associated with
that abstraction frame applies. Figure 2 shows the development. In this case, it is clear that the axis of system
sequential flow of time and system development. The abstraction is the physical axis. The system relationship
system components from the earlier abstraction frames are from the top down can be defined as “is composed of”,
used in the development of the systems in the later while the system relationship from the bottom up is given
abstraction frames. as “composes.” A fairly small set of standard, system
abstraction frames would be able to cover most
Sequence Sequence Sequence Sequence fundamental driving system relationships and abstraction
Frame (n-1) Frame (n) Frame (n+1) Frame (n+2)
axes in a comprehensive manner. The development and
standardization of a common set of system abstraction
frames would greatly enhance the communication of clear,
precise system information.

2 System Construction
Given the previously defined concepts in this paper, a
system or system of systems must be constructed in a single
abstraction frame. The fundamental boundaries of an
abstraction frame are time and space, which makes the
∆T n-1 ∆T n ∆T n+1 ∆T n+2 abstraction frame highly definable. A system production
schedule is an example of defining a series of times and
Figure 2 Abstraction Frames contexts that represent specific system abstraction frames.
The identification and communication of the controlling
system abstraction stacks is much more difficult than the
1.2 Abstraction Stacks
identification and control of system abstraction frames.
While a large complex system and/or system-of-systems
A key aspect of system and/or system-of-systems design
development will have only one set of abstraction frames
is the identification and communication of characteristics
(encoded in the program schedule), there may be hundreds
and attributes at the system-element level. In addition, the
of system abstraction frames that are recognized by only
system context and the general relationship that relates
specific groups or components of the system development
these system elements between different levels of system
organization.
abstraction must also be identified and communicated.
The classical systems engineering approach controls the
Abstraction stacks have been developed to facilitate the
system product development using an integrated master
communication of this type of structured system
schedule as well as strong requirements control and
information [4]. Figure 3 provides a representation of a
configuration management control. However, in the case of
“house system” that is abstracted along the physical axis of
system-of-systems development, this may not be the case.
abstraction.
For certain types of system-of-systems development, it may
never be possible to apply the level of technical
House management control that is commonly applied in the
classical systems engineering approach.
United States electrical power system deregulation and
Rooms associated system failures will be used as a first discussion
example. The backbone of the United States electrical
Floors, Walls, power grid will be represented by two types of power
Ceilings production facilities: hydroelectric and nuclear power
plants. While the main components of the United States
Lumber, hydroelectric system were developed during the 1930’s and
Bricks 1940’s, the nuclear power plants were mainly developed in
the 1960’s to 1980’s. The major programs that developed
the hydroelectric production were government programs
Molecules that used a top down, central control process to design,
build and manage these power systems as government
Figure 3 Abstraction Stack controlled utilities. Most of the nuclear power plants were
The representation of system behavior and structure is a designed and developed by either public or private
standard tool in many specialty areas of systems design and corporations to produce and sell power to meet customer
integration. There is no common convention or set of demand. A large-scale, static power grid was designed to
abstraction frames that have been established to clearly distribute the power to industries and customers in
communicate the semantics of the system design and populated areas.
 Over a period of time, the power production, well developed. Over the period of time that the large scale
distribution, and consumption systems were operationally system of systems was developed, the environment in
integrated into a system of systems capable of providing for which these systems were deployed has changed. These
the power needs of the United States. Then one controlling changes include the connection of industrial control nodes
aspect of this system of systems changed. This change was to computing systems that are reachable from the open
associated with how power was valued and priced. A Internet. Another change in the environment is the
power market was developed to buy and sell power to the development of very effective computer node and network
highest bidder. There was no basic technology change node attack procedures and processes. While under initial
associated with the major system-of-systems change, only a environmental conditions, the network control system
change in the controlling abstraction stack. The original additions were effective and appropriate. As the system
static electrical power distribution system was unable to environment changed, these network control nodes
dynamically adapt to these changes in demand. The provided a system attack vector that is open to anyone on
electrical power market idea was developed at a high level the Internet.
of abstraction. The ability of the power grid to deliver the
produced power to where it was needed was a given part of 2.2 System Disruption and Adaptability
this high level system abstraction. Therefore, the difference
between the age of the basic system components and the Another area of frequent large-scale, system-of-systems
controlling system abstraction contributed to system failure. failure is found in the business arena. One specific pattern
This is an example of a system-of-system failure based on associated with technology, system and system-of-system
the large time differential between the design and development that contributes substantially to business
deployment of the basic system building blocks and the failure has been identified as disruptive technology
controlling system-of-system operational concepts. The development [5]. Disruptive technologies start as small
controlling system-of-system attributes can be categorized scale, niche types of technologies that are too small or
using abstraction frames and abstraction stacks in a manner insignificant to be addressed by companies that are market
that highlights system configuration aspects that increase leaders in the product and associated technology areas.
the risk of system failure. Companies that are market leaders have value systems and
management styles that prevent their interaction with, and
2.1 System Modification adoption of, these disruptive technologies. This fact creates
an opportunity for other, more agile and adaptive
The power distribution system example will now be companies to exploit the disruptive technology to their
expanded to facilitate the discussion of another example of strong market advantage.
system-of-system failure potential. Approximately 50 years One key example of a disruptive technology is that of
after the development of the hydroelectric system in the open source software products. The extensive availability
United States a revolution occurred in computing of open source software products distributed by open source
technology, networking technology and large-scale system software companies is challenging the market leaders in the
monitoring and control. Unlike the centrally planned, top- closed source software field. Some established companies
down government approach that was used to develop the like the Java Company (Sun Microsystems) and Microsoft
basic components of the power production system, the have been strongly impacted by open source software.
computer and networking revolution was motivated by Other companies like IBM have quickly adapted their
private profit, capturing market share, and being a first product lines to take advantage of open source software
mover in these developing market areas. These motivations products, and have seen less of an impact from disruptive
combined with over 20 years of explosive physical open source software technology.
infrastructure and system architectural growth has created A key observation from the disruptive technology
an integrated set of system of systems that are not naturally example relates not to the technology or system component
secure. type, but directly to the value assigned to the technology by
The great value achieved by implementing the initial a specific organization. As more organizations acquire and
system control and monitoring mechanisms was not use a given disruptive technology, the value position of the
balanced against the future security and liability costs. This technology becomes greater and applies more pressure on
cost and benefit system evaluation was not performed for a the existing market leaders to engage the disruptive
few basic reasons. One primary reason is the fact that the technology. Therefore, the first key attribute required for a
future state of any system and global impact of any large- successful system-of-systems deployment is adaptability.
scale system is difficult, if not impossible, to predict. System flexibility is a related and important system
Another reason for the absence of a complete cost and characteristic that is viewed as the ability to adjust to a
benefit analysis lies in the fundamental manner in which the predictable range of environmental changes. System
value proposition of a large-scale, networked system is adaptability is a system characteristic that allows a system
developed over a period of time. When a network is small, to adapt to unknown and unforeseen environmental
it is a lower value target than when the network is large and situations, and is a key primary attribute of successful,
complex system of systems. An adaptable, flexible system changing nature of the system environment, its components
of systems must then also have the capability to sense and and system-of-systems relationships. A system that is a
interact with its environment. This sensing and interactive component of a system of systems controls the connection
capability is usually associated with a layered value system to the system-of-systems interface. The individual system
and an intelligent control mechanism. must determine that being a member of the larger system of
systems provides more individual system benefits than the
2.3 System Adaptation Interfaces cost of connection and/or membership in the system of
systems.
If uncertain future environmental conditions are a large The ability to make and control these types of
source of system failure, and adaptation is a key property of “connection - no-connection” decisions and actions is a
a system and/or system of systems that allow the system to fundamental system-of-system component capability.
effectively cope with an uncertain future, how should a While an individual system is usually organized around a
system be designed to take full advantage of the benefits of small set of controlling system relationships combined with
adaptation? A layered, modular system architecture that a well defined set of abstraction frames, the system
exposes multiple system interfaces is proposed as a design relationships and abstraction frame span can vary widely in
mechanism that supports these needed system a system-of-systems configuration. For example, open
characteristics and attributes. markets for capital and resources are artificial constructs
A system using the IBM 4758 Common Cryptographic used by large distributed groups of people to value
Architecture as a key management component in a products, services and resources. A value is also assigned
hardware security module will be used as an example of an to the liquidity of these items in the market over any given
adaptable interface. A security component of Automatic period of time.
Teller Machines (ATM) was deployed in a hardware As discussed earlier in this paper, a system of systems
configuration. A very large number of the ATM systems in that is designed based on the capacity of the physical
the world use this security component. After the system constrained by laws of nature, may fail dramatically
deployment and use of the ATM systems, a small number when artificial laws are applied to the evaluation of system
of successful system security attacks were perfected. The connection, control and operation. The fundamental
details of these successful attacks spread quickly across the differences between the natural sciences, and sciences of
Internet. The changed system environment created a high the artificial, creates an area in complex system design,
risk for anyone using this security component to protect and deployment and operation that has a very high potential for
control the distribution of funds. generating decisions and actions that directly relate to
Addressing this new security threat was constrained by system-of-system failure.
the number of deployed systems and the prohibition against Natural science is based on the pursuit of knowledge and
changing any part of the currently deployed physical the discovery, documentation, and communication of the
hardware security system. The key to the successful basic unchanging natural laws of the universe. The
solution of this problem was found in the interface between scientific method has been used to create a large body of
the user and the security hardware component. Due to scientific knowledge in a top-down fashion. Knowledge
modular system architecture and well-defined, well- and scientific theory were developed and validated at one
documented system interfaces, it was possible to insert level of abstraction, without complete understanding,
another system security component between the user and visibility and insight into the layer of knowledge of the next
the legacy security component. In the case referenced here, level down in the abstraction stack. Houses were built from
an artificial neural network approach was proposed to wooden timbers before the detailed calculations of force
implement an additional security policy layer that addresses and energy were perfected to enable the practice of modern
the well defined set of successful security attacks [6]. material science and structural engineering. Structural
The insertion of the artificial neural network based engineering was developed and applied before the material
security policy component at the interface between the user molecular and atomic forces were completely understood.
and the static legacy hardware component is an example of And science continues to advance as individuals continue to
adapting a fixed, static system by adding an adaptable, search for, and perfect, the “Theory of Everything” that will
flexible component. This component insertion was directly describe and communicate the basic unchanging laws of
dependent on existing modular system design and nature [7].
accessible interfaces. On the other hand, a system of systems is directly
dependent on knowledge of the next lower levels of system
physical components and/or abstraction stacks to determine
2.4 System Environmental Awareness when, and the degree to which, the current system-of-
system configuration will operate successfully. Further, the
The ability to detect, track and understand the changes in span of the required knowledge may be evaluated by first
the system environment is a valuable characteristic of a deciding whether the operational science and technology
system. In a dynamic system of systems, the environmental lies in the realm of natural or artificial sciences. This initial
awareness ability is even more critical because of the
evaluation of the system of systems is used to guide further References
evaluation and investigation into the system attributes and
characteristics that will increase the probability of the [1] Joseph J. Simpson, A Generic, Adaptive Systems
development of a successful system-of-systems deployment Engineering Information Model, Missouri University of
and operation. Science and Technology Thesis, Rolla, Missouri, 2004.

3 Summary [2] J.J. Simpson and M.J. Simpson, “Systems and

Objects,” Proceedings CD, Thirteenth Annual International
The characteristics and attributes of successful systems Symposium of the International Council on Systems
and system of systems have been analyzed and evaluated in Engineering, Crystal City, July 2003.
this paper. System abstraction frames and system
abstraction stacks have been presented to facilitate the [3] J.J. Simpson and M.J. Simpson, “System Integration
organization and communication of system context Frameworks,” Proceedings CD, Fifteenth Annual
information essential to the successful deployment of International Symposium of the International Council on
system of systems. The following key system attributes and Systems Engineering, Rochester, July 2005.
characteristics have been identified as essential components
of successful systems: flexibility, adaptability, modular [4] J.J. Simpson and M.J. Simpson, “Formal Systems
design, open interfaces, and contextual awareness as well as Concepts,” Proceedings CD, Fourth Annual Conference on
local system control over connection to global system-of- Systems Engineering Research, Los Angeles, April 2006.
systems resources. Adaptable system interfaces are
essential in the construction of system of systems that have [5] Clayton M. Christensen, The Innovator’s Dilemma,
the ability to adjust to uncertain future environments. Harper Business, Boston, Massachusetts, 1997.
System interfaces are considered as the primary source
of system complexity by several authors. Casti provides an [6] R.T. Gordon and A. Gehrke, “Controlling Security
analytical treatment of system complexity, providing a base API Attacks: An Artificial Neural Network Approach,”
model of system complexity that assigns the source of Proceedings of the Artificial Neural Networks in
complexity to the interface between interacting systems. Engineering Conference (ANNIE 2007), St. Louis,
Casti asserts that the magnitude of the complexity generated November, 2007.
by the system interface is a functional value established by
one of the systems that participates in the interface [8]. [7] Herbert A. Simon, The Sciences of the Artificial, MIT
Warfield also assigns the source of complexity to the Press, Cambridge, Massachusetts, 1996.
interface and interaction between systems. Warfield details
a scientific approach to system design that is focused on [8] John L. Casti, Alternate Realities, Mathematical
reducing system complexity and large scale system failure Models of Nature and Man, John Wiley & Sons, New York,
[9]. Clearly, a critical aspect in the deployment of New York, 1989.
successful system of systems is the ability of the system
architect and engineers to provide a controllable system [9] John N. Warfield, A Science of Generic Design:
design that reduces the complexity associated with the Managing Complexity Through Systems Design, Iowa State
system as its interfaces interact with a wide range of other University Press, Ames, Iowa, 1994
systems, system components and the system’s environment.

4 Conclusions
The power of a large scale system of systems can be
realized given the proper system design and operational
context. While many of the general characteristics and
attributes of successful systems were identified in this
paper, further work needs to be accomplished to categorize
and evaluate the possible types of system operational
environments for any future system. Highlighting the
difference between the controlling aspects of natural
science and the operational aspects of the artificial sciences
in these analysis and evaluation activities will provide
further insight into the proper balance for system
architectural features.