ultra large scale systems design using list
TRANSCRIPT
-
8/14/2019 Ultra Large Scale Systems Design Using LIST
1/16
Designing Ultra Large Scale Systems
Can Lean Inventive Systems Thinking (LIST) Help?
Navneet Bhushan and Karthikeyan Iyer
Crafitti Consulting Pvt Ltd, (www.crafitti.com)Emails:[email protected],[email protected]
1B-401, Akme Harmony,Sarjapur Outer Ring Road,
Bangalore 560037, INDIA
Abstract
The challenges of designing the needed Ultra Large Scale (ULS) systems are beyond
the methods and techniques that humanity currently knows of. These systems are
characterized by extraordinary decentralization, inherently conflicting, unknowable and
diverse requirements, continuous evolution and deployment, heterogeneous,inconsistent, and changing elements, erosion of people/system boundary, normal failures
and new paradigms for acquisition and policy. Today the largest systems being designed
are what in the US military parlance are called System of Systems (SoS). The current
cutting-edge SoS are characterized by operational and managerial independence of
elements, evolutionary development, emergent behavior and geographic distribution.
ULS will be SoS at Internet Scale. This is not a simple matter of extending the current
approaches as P.W. Anderson in his 1972 classic paper described More is Different. We
have definitely come a long way from that time in our approaches to design systems. Yet
predominantly our approaches continues to be constrained by the analytical and logical
thinking (analogical thinking) that we have perfected over past centuries.
The research agenda proposed to design ULS systems includes Human Interaction,
Computational Emergence, Design of all levels, Computational Engineering, Adaptive
System Infrastructure, Adaptable and Predictable System Quality, Policy, Acquisition and
Management. In this paper we explore the suitability of different thinking dimensions for
designing ULS Systems these are Lean Thinking, Inventive Thinking and Systems
Thinking. Our hypothesis is that these thinking dimensions need to play a much larger
part than the current analogical thinking that we are used to, in order to design these
highly complex systems of the future. We propose our framework named the Lean
Inventive Systems Thinking (LIST) as a possible approach to design such ULS Systems.
Keywords: Ultra Large Scale Systems, Systems Thinking, Inventive Thinking, Lean
Thinking, Lean Inventive Systems Thinking.
1. IntroductionThe ability to reduce everything to simple fundamental laws does not imply the ability to start from
those laws and reconstruct the universe says P.W. Anderson in his classic paper titled More is
Different[1]. He further states, The constructionist hypothesis breaks down when confronted with the
http://www.crafitti.com/mailto:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]://www.crafitti.com/ -
8/14/2019 Ultra Large Scale Systems Design Using LIST
2/16
twin difficulties of scale and complexity. The behavior of large and complex aggregates of elementary
particles, it turns out, is not to be understood in terms of simple extrapolation of the properties of a few
particles. Instead, at each level of complexity entirely new properties appear, and the understanding of
the new behaviors requires research (fundamental). We are standing at an important point in our
history when this century of complexity will lead to extremely large scales systems designed by humans.
The scale is the new frontier. These systems called the Ultra Large Scale (ULS) [2] systems, demands
unprecedented capabilities from human minds to design, operate, control and manage these systems.
The ULS systems are characterized by increasing interactions, dependencies, couplings or connections in
not only the depth of existing dimensions but in increasing the number of connection dimensions
manifold. The complexity of the ULS systems will be increasing manifold as it demonstrates the complex
behavior associated with complex systems such as a natural eco-system. Dealing with this level of
complexity needs new methods of study or solving problems. These methods need to be not only aware
of but thrive and exploit very nature of complexity. This nature is characterized by indeterminacy, non-
linearity, chaos, adaptation, self-organization and distributed intelligence [3]. The connotation of difficult
or hard to describe leads to a system being viewed as complex. Further complexity can be considered asa contradiction of distinction and connection [3].
Complexity has been defined as two or more distinct parts that are joined in such a way that it is difficult
to separate them. This characteristic of complex system led Ray Kurzweil [4] to state I know English but
none of my neurons do. Complexity generates an emergent behavior that is not exhibited individually by
any of the parts, partially or fully. Further the connotation of difficulty in understanding this emergent
behavior of complex systems springs from our classical methods of thinking [3]. These methods have
proven their worth in tackling the issues of classical science, problems and issues predominantly in a
world of distinct organization structures where the interactions between uniquely identifiable elements
were clear and unambiguous, relatively few and limited to specific dimensions. These methods werebased on reductionism or analysis, determinism, dualism, correspondence theory of knowledge and
rationality. However, the classical methods of dealing with mechanical systems and mechanistic
worldview pioneered by Aristotle and taken to their zenith by Newton, faltered in explaining the new
observations that started coming in early parts of last century. Further, the lessons from complexity
research initiated a worldview that real world is not the perfect, geometrical, ordered, predictable,
deterministic, rational construct that human mind, labor and ingenuity has created by engineering
perfect geometries that we see in all man-made physical structures. The nature turned out to be an
extremely creative and complex system, where dynamism and emergence are the norms. It is with
deeper study of nature of information; man started realizing the inherent complexity of the world that is
unfolding.
Our research indicates that the successful thinking dimensions that are working successfully in dealing
with complex systems in contrast to the classical analogical thinking are Lean thinking, Inventive thinking
and Systems thinking. We propose a framework for design of ULS systems using the integrated Lean
Inventive Systems Thinking (LIST). The paper is organized into following sections. In Section 2, we briefly
describe the challenges and needs of Ultra Large Scale Systems design. In Section 3, we describe
elements of design coming from Lean Inventive Systems Thinking. In Section 4, possible approaches for
-
8/14/2019 Ultra Large Scale Systems Design Using LIST
3/16
Ultra Large Scale Systems Design as emerging from LIST thinking are described. The paper ends with
conclusions and details of future research directions in Section 5.
2. Ultra Large Scale Systems Challenges and NeedsGiven the systems that we have built and which are continuing to scale-up in all walks of life, we arecloser to building larger and larger systems. There are needs for such systems to optimally utilize the
rapidly depleting natural resources and also to function in a highly connected world that we have created
for ourselves. Most of these systems are, be it web and computing infrastructure, supply chain systems,
healthcare infrastructure, military systems or government systems, software based engineering systems.
These systems are increasingly complex web of ultra-large, network-centric, real-time, cyber-physical-
social systems.
The ULS Systems will be system of systems at the Internet scale. Characteristics of ULS systems arise
because of their scale. These are unprecedented decentralization; inherently conflicting, unknowable,
and diverse requirements; continuous evolution and deployment; heterogeneous, inconsistent, andchanging elements; erosion of the people/system boundary; normal failures and new paradigms for
acquisition and policy. Table 1 gives a brief view of contrasts between present approaches and
characteristics of ULS Systems [6].
Table 1: Contrasting the ULS Systems needs and Current Approaches
ULS Characteristics Present Approaches
Decentralized Control All conflicts must be resolved and resolved centrally and
uniformly
Inherently conflicting,
unknowable, and diverserequirements
Requirements can be known in advance and change slowly.
Tradeoff decisions will be stable.
Continuous evolution and
deployment
System improvements are introduced at discrete intervals.
Heterogeneous, inconsistent,
and changing elements
Effect of a change can be predicted sufficiently well.
Configuration information is accurate and can be tightly
controlled. Components and users are fairly homogeneous.
Erosion of the
people/system boundary
People are just users of the system. Collective behavior of people
is not of interest. Social interactions are not relevant.
Normal Failures Failures will occur infrequently. Defects can be removed.
New paradigms foracquisition and policy
A prime contractor is responsible for system development,operation, and evolution.
The ULS systems will be artificial systems hence they differ from natural complex systems in fundamental
ways. Unlike natural systems that may evolve because of specific constraints or available paths, the
artificial systems are designed at least in principle, with a specific goal or function in mind. As Herbert
-
8/14/2019 Ultra Large Scale Systems Design Using LIST
4/16
Simon [5] describes, an artifact is an interface between inner environment and the outer environment.
The artifact tries to accomplish a goal or provide a function in the outer environment. This artifact can
have one of many possible internal environments to accomplish the same desired function in the same
environment. This is an important fact, as it indicates that theoretically infinite ways exist to construct or
design an artifact to accomplish specific function in specific environment. This is important; because this
fact creates an uncertainty and unpredictability, in the artificial world that we are living in as it leads
different actors to design different artifacts to achieve the specific function in multiple environments.
This is a dimension of complexity that needs to be understood and grappled with.
The interplay of natural and artificial is another area that comes under the realms of ULS scale
complexity. Natural objects evolve through natural selection and based on the environment in which they
operate. The observations based on how the natural phenomena occur led humans to fields of natural
sciences. The industrial revolution started a focused direction towards the artifact sciences where
suddenly man-made objects became prevalent and useful with specific functions or goals to be achieved
in specific environments. Modern world characterized by artificial environments, virtual reality and
synthetic materials, has become more man-made than natural. Yet nature has not been tamed fully infact natures fury keeps on giving clear messages of the journeys that humankind has yet to perform, in
the form of earthquakes, hurricanes, floods, volcanic eruptions and multiple natural disasters that
happen in many parts of globe.
The ULS Systems research needed as described in [2] include 7 main fields. These are represented in
Table 2.
Table 2 Ultra Large Scale Systems Research Areas
ULS Systems Research Area Specific Sub-Areas
Human Interaction
Context-Aware Assistive computing Understanding Users and Their Contexts
Modeling Users and User Communities
Fostering Non-Competitive Social Collaboration
Longevity
Computational Emergence Algorithmic Mechanism Design
Metaheuristics in Software Engineering
Digital Evolution
Design
Design of All Levels
Design Spaces and Design rules
Harnessing Economics to Promote Good Design
Design Representation and Analysis
Assimilation Determining and Managing Requirements
Computational Engineering
Expressive Representation Languages
Scaled-Up Specification, Verification, and Certification
Computational Engineering for Analysis and Design
Adaptive System Infrastructure
Decentralized Production Management
View-Based Evolution
Evolutionary Configuration and Deployment
In Situ Control and Adaptation
-
8/14/2019 Ultra Large Scale Systems Design Using LIST
5/16
Adaptable and Predictable System
Quality
Robustness, Adaptation, and Quality Attributes
Scale and Composition of Quality Attributes
Understanding People-Centric Quality Attributes
Enforcing Quality Requirements
Security, Trust, and Resiliency
Engineering Management at Ultra-Large Scales
Policy, Acquisition, and
Management
Policy Definition for ULS Systems Fast Acquisition for ULS Systems
Management of ULS Systems
The increasing complexity either natural, artificial or a combination of both is an important fact of the
new world. The increasing scale of ULS scale systems is creating more complexity as is evident in multiple
dependencies, connections and unknown network effects the new pace is creating for humankind. Yet
there is a hope it is remarkable how much the human mind has been able to create and synthesize
especially in last 100 years or so. In fact, there is no gainsaying that artificial future is more likely than a
natural future at least next 50 years or so. This is possible and thinkable only because of the human
mind which has proven to be an extremely robust and comprehensive factory of new ideas, new
thoughts and new memes that are successfully implemented to create artifacts, synthetic environments,
and robust global artificial systems. This is the dimension of design thinking. In the next section we will
describe elements of design that we have culled out from Lean Inventive Systems thinking.
3. The Elements of Ultra Large Scale Systems Design A SystemsApproach
We do not understand complexity. This is an inherent property of reality and the nature shows umpteen
examples of the complex behavior as exemplified in the emergence of order in various seemingly simple,
local interactions with limited rules in various natural systems. Complexity emerges from dependencies
informational, control, decisions, structural and material dependencies. Connections create complexities
as well. Multi-dimensional dependencies create higher order complexities. How can one embrace
complexity rather than thinking or working for reducing complexity? How can one invent or innovate to
leverage the boundaries rather than always focusing on the core? It is at the boundaries that value
capture will have maximum returns. Value capturing and value creation are interrelated phenomenon
that one needs to maximize. Further complexity is not necessarily natural; in fact most prevalent form of
complexity impacting human-beings is artificial. The artificial things are synthesized by human beings.
The world around us is full of more artificial things than natural things. Starting from our morning alarms,
newspapers, radio music, television images and computer chat rooms we work through artificial things
till the night when we sleep in artificially heated rooms and beds.
-
8/14/2019 Ultra Large Scale Systems Design Using LIST
6/16
The very purpose of design is to reduce the complexity of what is abstract, by giving it some sort of
physical shape or form. The crystallization of thoughts into some structure, either on paper or on any
medium reduces complexity as perceived by the human mind structures make it easy for the brain to
memorize and retrieve information. By design and not by accident highlights the scientific, structured
aspect of design, where the system, its components and its behavior are completely understood, thereby
introducing the element of predictability into the proceedings. Ultra large scale systems design needs a
holistic systems approach incorporating the following key elements:
3.1.NeedsWhat the users of a product need from it is probably the most ambiguous question of all. User needs
stem from multiple intelligences
[19], namely linguistic, musical,
logical, spatial, kinesthetic,
inter-personal and intra-
personal. User needs also stem
from the gratification of the five
senses sight, sound, touch,
taste and smell. At product-user
interface points, the product has
the opportunity to beneficiallyimpact one or more senses or
intelligences. For instance, one
could look at the interface
points on a product use timeline
or life cycle similar to the Buyer
-
8/14/2019 Ultra Large Scale Systems Design Using LIST
7/16
Utility Map [18] and design the product to raise the bar of user experience (intelligences + senses) at each
of the interface points.
Simultaneously, one
could look at the spatial
and qualitative interface
points.
User needs may also
vary over time or due to
change in context. This
leads to a multi-
contextual and evolving
map of needs. A multi-
level user needs
dashboard can serve as
a key tool in designingthe next level of user
experience.
3.2.Function (Inventive Thinking)Needs and functions are often interchangeably
used. It is common to hear that the function of a
product is to fulfill a certain need e.g. the function
of a washing machine is to fulfill the need of
cleaning clothes. While related to each other,
needs and functions differ in perspective an
emotional human-centric view vs. a parametric,
neutral view. One could argue that the function of
a washing machine is to wash clothes, not to
fulfill the need to clean clothes.
A single product may have several functions which
may be sorted based on priority or importance.Typically, one core function is enabled or
supported by supplementary functions. There may
be additional layers of derived/ dependent
functions.
Products can be synergistic when the cores attract (they are different from each other) and the outer
functional layers merge or fit well into each other. On the other hand, two products with repulsive cores
-
8/14/2019 Ultra Large Scale Systems Design Using LIST
8/16
(very similar to each other) will typically compete or in the unlikely event of a forced merger try to
subsume each other.
Function-focused design looks at two aspects:
1. Performing the function elegantly [7] (in the simplest manner possible minimal consumption ofresources and minimal harmful effects) [46][47]
2. Performing only what is necessary (eliminate unnecessary functions) [8]At the outset, there may be multiple design alternatives or paths, each offering a functional improvement
and therefore looking equally promising. How do we know which path to take? TRIZ offers clues and
directions.
TRIZ is a large collection of empirical methods discovered and invented through comprehensive studies of
millions of Patents and other inventions for problem formulation and possible solution directions. Reader
can refer to large body of knowledge at [46][47]. TRIZ clearly distinguishes two main parts of problem
solving problem description/definition and its solution.
Define, Describe, Analyze the problem from multiple perspectives, as deep and as wide as onecan go. This requires a focused discipline to not to jump to solution immediately TRIZ has
tools and Processes for Problem Definition
Find out the root contradiction and look at how the contradiction has been solved in the past Solve by exploring in multiple directions but start from the end result The Ideal Final Result
Focus on Functionality not features.
Some of the key design directions that emerge from TRIZ are:
1. Choose a design which eliminates one or more system contradictions (contradictions are pointswhere beneficial impact to one parameter causes a negative impact to another parameter)
2. Choose a design which moves the system to a higher level along known lines of systemevolution.
3. Choose a design which is moving towards ideality all benefits at zero cost.3.3.StructureOften, form (or structure) determines function,
the way the function will be achieved and the
constraints that the function will face. Structures
provide stability to systems, preventing descent
into randomness or chaos. The flip side is that
stability brings rigidity along, suppressing the
ability of a system to change, adapt and evolve.
Can structures overcome this basic contradiction
provide flexibility while retaining stability?
-
8/14/2019 Ultra Large Scale Systems Design Using LIST
9/16
Structures also have to evolve in step with needs and functions in order to enable systems to move to the
next level. The TRIZ lines of evolution also clearly point out the significance of increasing structural
flexibility to overall system evolution.
Why are specific system elements rigid?
a. Stability - They are elements where the most critical functions are performed and cannotafford to fail
b. Unpredictability They are not very well understood and are unpredictable, hence they aredeliberately operating within strict constraints
c. Dependency Too many other system elements are dependent on this element, hence itcannot be changed very easily and without pain
d. Insulation They are not very well connected and therefore do not have an incentive tochange or adapt to changes
e. Efficiency The elements are optimally structured to perform certain functions as efficientlyas possible
All of these reasons are fundamentally linked to the concept of system complexity as defined by coupling-
cohesion measures [32]. Tightly coupled structures tend to be rigid and therefore not amenable to
change. Cohesive system elements and loosely couple structures are conducive to change, without
compromising on stability. The root cause of high coupling and low cohesion is the lack of understanding
of structural relationships between system elements, the need for these relationships, when they are
activated/ de-activated and the strength of these relationships. The following steps are useful to move
towards increasing structural flexibility:
1. Find the areas of the system that are highly complex (tightly coupled, non cohesive) usingDependency Structure Matrices [48] and the System Complexity Estimator [12].
2. Continuously re-architect the system to decrease coupling and increase cohesion.Another critical aspect of structure is the concept of centre centre of gravity or centre of control.
Centralized structures tend to have very tightly coupled central nodes or hubs. While centralized
structures offer a lot of stability, the hubs tend to become bottlenecks in complex, quickly changing
environments. Recent trends of system evolution seem to be pointing towards networked, distributed
architectures with distributed centers of gravity and distributed intelligence.
3.4.BehaviorWhile needs, functions and structure characterize the static system, the working system is characterized
by its behavior manifested in the way system elements come together to create a whole greater than the
sum of the parts. In real time, system elements interact between themselves and with external entities,
reacting differently to different stimuli and evolving in the process.
Human beings have typically designed systems that are based on centralized intelligence. This can
probably be attributed to the limitations of the brain to handle complexity [43]. On the other hand,
natural systems are evidently much more complex and display what we perceive as emergent behavior.
-
8/14/2019 Ultra Large Scale Systems Design Using LIST
10/16
They are so named because we are unable to draw logical threads from any central intelligence to the
behavior displayed by the elements of the system (in other words, behavior that cannot be explained by
logic). As systems grow larger, the cost of centralized control increases. Gradually, multiple centers of
intelligence emerge and finally, intelligence is so completely distributed that hubs cannot be easily
isolated.
While emergent behavior has been observed, it has never been designed. In order to design
emergence, the characteristics of emergent behavior and systems displaying such behavior have to be
understood in detail. Following are some of the key characteristics:
Synchronization
All systems (animate and inanimate) seem to have the innate tendency or ability to synchronize with the
surroundings [20]. In the absence of centralized control, there still seems to be a strong pull for elements
to work in tandem. Intuitively, it makes sense. Much more can be achieved with much less effort through
synchronized efforts. While, the initial belief was that this behavior is the result of complex intelligence as
pursued by game theory [22], the evidence of synchronized behavior in entities with very low intelligence
(fire flies) or with no intelligence at all (pendulums) has changed the assumptions somewhat. In small
systems, stable behavior can be achieved through stable structures and centralized logic. In large,
complex, non-linear systems, these measures are typically inadequate or ineffective. Observations from
natural system behavior show that the ability of system elements to synchronize (on their own) helps
maintain stability and prevents the system from collapsing (natural tendency to decay and degenerate
into chaos). While most designs are built with the objective of overcoming entropy through tighter
centralized control, natural systems balance the tendency of systems to move towards higher entropy
with the counter tendency to synchronize. There are clear parallels to be drawn with USAs war against
terror and potential alternative designs to combat terror.
How to design complex artificial systems with built-in ability to synchronize?
1. Stop trying to build more and more complex central brains. Distribute the intelligence.2. Even in highly complex natural systems, synchronization happens through simple rules of state
change or behavior change. These are preset responses to stimuli and may even happen without
any active intelligence. Simpler elements are easier to influence (social network theory) and
therefore contribute to synchronization much more effectively.
3. Sync happens around rhythms (or clocks). Clocks do not have to be centralized, but central clockscan play a major role in assisting synchronization (touched upon in greater detail in the chaos
section). The Takt time [8] concept also seems to be alluding to something similar.
4. There needs to be an innate reason to synchronize (although it is not essential for all elements toknow the reason). Sometimes, reason is so deeply entrenched in behavior, it ceases to be visible;
this can be dangerous (some rhythms are just impossible to get out of). If synchronization offers
benefits in reaching some common output, elements tend to synchronize over multiple iterations
of discovery.
5. All system elements need to be designed as learning systems. Synchronization has to be learnt;the time period of learning may vary.
-
8/14/2019 Ultra Large Scale Systems Design Using LIST
11/16
Chaos
Chaotic systems look chaotic at the outset the patterns of behavior are non-linear but not random [21].
They hide certain patterns of behavior called attractors. Attractors are essentially axes of stability,
invisible at the outset but clearly borne out by macro behavior. Again, while natural systems are known
to display chaotic behavior, artificial systems are not typically designed for chaos. There are certain
traits that can be exploited for such design:
1. In chaotic systems, attractors manifest across scale (clearly, if they do not, system stability willget affected). In reverse, while designing large, complex systems, all elements need to gravitate
towards a common attractor (the behavior that needs to be manifested). For example, in a
complex, non-linear system such as the internet, if security is a characteristic that needs to be
implemented, it cannot be done by inserting some central security nodes into the network.
Secure behavior needs to manifest across scale (entire network, zones, local networks, smallest
subnet, router within a subnet, TCP/IP socket on a router etc.)
2. Attractors act as strong central rhythms or clocks. You can choose to design systems such thatthe central rhythm is beneficial to the objectives of the system. In the absence of such design,emergent attractors can often turn out to be out of sync with the system objectives (many
large organizations face this challenge when they grow in size and changes become more and
more difficult to incorporate [51]).
Balance
In the world of complex, chaotic systems, stability is not about being still rather it is about being on the
move constantly to avoid stagnation and to counteract the inevitable forces that prevent you from
remaining still. (Ever tried to balance a cricket bat on one of your fingers?) The key element is not
stillness but balance. For instance at a macro level, system behavior may be looked at as a function of
the following four elements and their individual traits:
1. Fire fuel, energy to run the system2. Water vitality or life of the system, ideas3. Air transports essentials to all parts of the system, enables change4. Earth structure and raw material, the basis of the systemIn centralized systems, specific parts of the system may be designated specific roles in alignment with
one of the four elements. The overall proportion of the elements is also controlled centrally and changes
over time. In complex systems, these elements stay in overall balance, present at all levels ranging from
micro to macro but in different proportions depending on the system need.
This is very close to Ayurveda [52] which prescribes mechanisms to sustain balance between the four
elements (as manifested in the human body as a system) and to prevent and cure imbalance, if any.
Consider an organization as a complex entity, sustained by motivators, thinkers, communicators and the
implementers representing fire, water, air and earth respectively. While some portions of the
organization will specifically need larger proportions of certain types (e.g. leadership, research, sales and
-
8/14/2019 Ultra Large Scale Systems Design Using LIST
12/16
engineering), sustained stability and growth requires an overall balance to be maintained. Any excess or
shortfall in any of the elements proves detrimental in the long run.
The concept of balancing key elements of a system which are manifested across scale (again falling back
on the chaos attractor theory) can be used as an alternative, simpler mechanism to design system
behavior (as compared to typical sequential, deterministic approaches).
4. The Process of Ultra Large Scale Systems Design A Lean Set-basedApproach
It is fascinating to note that while there are possibly millions of different products out there, there are
just a handful of different approaches to the design process.
Typically, product design has followed a convergent
approach to arriving at the right design, in other
words, survival of the fittest. Based on data
available at any given point in time, the best
alternative out of multiple alternatives is chosen as
the to-go approach. This chosen design then gets
fine-tuned over time, using the same survival of the
fittest approach iteratively. This is similar to
traversing a unidirectional tree (from root to
branches) with a single player; choose the best
looking branch at every node and hope to make
your luck as you go.
Fixing on one approach out of many early on in the life cycle has some clear pitfalls:
1. Always reaching Local OptimaAs a design matures, more data and information tend to become available. Often, there is the
feeling If I had known this earlier, I might have taken a different path! Decisions based on
incomplete data tend to lead towards local optima in many cases much inferior to global
optima [16].
2. Creates a false sense of simplification of complexityThe decision to take one path somehow erases the existence of other paths from the mind,
whereas in reality they are still there. This artificial simplification often results in all resources
getting expended in one direction making it more and more difficult to retrace and take
corrective measures as we walk deeper down a path.
3. Introduces artificial delayBeing forced to choose the best alternative may result in procrastinationLet me wait for more
information before I choose the path to move ahead. While this delay in action (making way for
rigorous analysis) may look justifiable, much of the analysis is only hypothetical (since the data
-
8/14/2019 Ultra Large Scale Systems Design Using LIST
13/16
required for real analysis can only be obtained a few steps down the line) and therefore not of
much use.
4.1.Set-based Concurrent EngineeringSet-based concurrent engineering [24][25][26][27][28][29][30][31] is a product development technique
invented by Toyota, which focuses on collaboration between different departments. The aim is at shorterdevelopment times with an increased quality level by improving collaboration and by parallelizing parts
of development process. In the traditional point based approaches the teams select an initial design
option and work on quickly producing it however, the design gets
modified as new information, experiences and requirements emerge
thereby creating what is called Design Churn effect. In this
scenario, the product remains in development phase for very long
period as the chain reactions created by many modifications to initial
design lead to continuous refinement and an evolutionary design
that keeps on going. This is the result of early design convergence
and action-oriented approaches most companies and management
gurus prophesize. In contrast Toyotas SBCE advocates slow
convergence strategy. SBCE processes starts with large design
alternatives covering broad design spaces and then slowly converges
to a possible design by eliminating the weakest alternatives rather than choosing one best alternative.
It is a counter-intuitive approach and looks paradoxical to people trained in the traditional point based
approaches. Various sets of alternatives are taken ahead for all parts of the product and the weakest
ones are eliminated as we move in the product development life cycle.
SBCE leading to slow convergence seems like an inefficient and expensive way to develop products,
however, Toyota creates new automobiles faster than industry average with less effort. It has beentermed as the Second Toyota Paradox as more time spent in early phases of the product life cycle leads
to less time spent in the overall product life cycle [26].
Although SBCE is known for many years and many research publications have described the process, it
has not been picked up by many companies as principles are counter intuitive and in time and budget
constrained commercial organizations, it becomes very difficult to not to show one design quickly so as to
show the development project is on the right track to the top management. The information, decision,
design and organization complexity also increases as SBCE as a process requires strict discipline in
following the process by everyone as there is no central control, it creates a self-organizing system.
Further, the SBCE principles dont describe specific methods, techniques, tools or frameworks for
execution. It is this important gap that Inventive Thinking approaches (TRIZ, function-based approaches)
and Systems Thinking approaches (holistic understanding of needs, function, structure and behavior)
serve to fill.
5. Linking the design elements with the LIST framework
-
8/14/2019 Ultra Large Scale Systems Design Using LIST
14/16
SBCE Steps Specific Actions TRIZ and Other tools
Mapping the
Design Space
Describe user needs In case of multiple needs carry out needs
interdependency analysis
Find out key functions to be performed
Understand structural complexities Understand behavioral complexities Function dependency analysis to find out
interdependencies
Can some high level functions specific tostrengths of different teams be identified
Let each team explore the specifications, needs,functions independent of each other
Each team explore design tradeoffs throughsimulations and their past observations
Each team should come up with their sets ofdifferent solutions with in the functional and
performance needs of the product
Problem Formulation andAnalysis
Value Stream [49]
Ideal Final Result (IFR)Why-what hierarchyNine windowsDependency Structure
Matrix (DSM) [48]
Function/Attribute AnalysisSystem Complexity
Estimator (SCE) [12][16]
S curve analysis Vedic Inventive Principles
[50]
Contradictions Technical/Physical
Trends of evolutionStriving for
Conceptual
Robustness
(Functional Team
level)
Design should remain functional after variationsin its environment
Vulnerability of system to changes in theenvironment should be minimized
Modularized Design with standard components
IFR AFD/Subversion Analysis Robust Inventive System
Design (RISD) [10]
DSMIntegration by
Intersection
(System level)
How are the parts integrated to meet at thepoint that will be regarded best solution
Find out overlap of feasible design spaces foreach sub component
Decisions about eliminating the weak designs
Decision DependencyMatrices (DDM) [13][15]
Analytic Hierarchy Process(AHP) [15][45]
Technical Contradictions /Inventive Principles
Establish
Feasibility before
Commitment
Multiple concepts developed using prototypingsimulation
The infeasible ones will be rejected rest all willcontinue to be developed
Decision theoreticprinciples [28][29]
AHP Closer to IFR
Conflict Handling Cooperative Conflict handling Which solution is closer toIFR?
DDM AHP
6. Conclusions and Further ResearchThe challenges posed by the Ultra Large Scale (ULS) Systems design are immense. These challenges are
unprecedented as well. Further the traditional methods that we have perfected and that have served us
-
8/14/2019 Ultra Large Scale Systems Design Using LIST
15/16
for many centuries are failing in designing such large scale systems. Taking the cues from research
agenda proposed by the ULS systems community, we propose in this paper a framework combining lean,
inventive and systems thinking, is a possible route. The framework that we termed LIST has elements of
natural evolution, design thinking, holistic thinking and inventive thinking. We propose to carry out
further research and development of the methodology for designing highly complex ultra large scale
systems.
References
[1] Anderson, P.W., More is Different, Science, Vol. 177, No. 4047, Aug. 4, 1972, pp. 393-396.[2] Northrop L., Ultra-Large-Scale Systems The Software Challenge of the Future, SEI Report, June 2006.[3] Gershenson C. and Heylighen F., How Can We Think the Complex?, http://www.calresco.org/offline.htm
(accessed on 12 August 2006)
[4] Ray Kurzweil, The Singularity is NearWhen Humans Transcend Biology, Penguin, US, 2005.[5] Simon H. A., The Sciences of the Artificial. 3rd edition, The MIT Press, USA, 1996.[6] Northrop L., Scale Changes Everything, ICGSE, Brazil, 2006.[7] May Mathew E., The Elegant Solution Toyotas Formula for Mastering Innovation, Free Press, NY, 2007[8] Liker J.K, The Toyota Way: 14 Management Principles from the Worlds Greatest Manufacturer, McGraw-HillProfessional, 2004[9] Bhushan N., Set-Based Concurrent Engineering (SBCE) and TRIZ A Framework for Global Product
Development, Proceedings of TRIZCON 2007, Altshuller Institute, 2007
[10] Bhushan N, Robust Inventive Software Design A Framework Combining DSM, AHP and TRIZ, 7th DSMConference, Seattle, US, 2005.
[11] Bhushan N. and Iyer K., Strategic Project Management Priorities in Global Software Development Scenarios,Communicated to IEEE Software, 2007.
[12] Bhushan N., System Complexity Estimator Applications in Software Architecture, Design and Project Planning,3rd International Conference on Quality, Reliability, Infocom Technology, ICQRIT, 2-4 Dec 2006, New Delhi,
India.
[13] Bhushan N., Decision Dependency Matrices, 8th International DSM Conference, Seattle, US, Oct 2006[14] Bhushan N., Set Based Concurrent Engineering Using DSM A Framework for Software Development, 8th
International DSM Conference, Seattle, US, Oct 2006.[15] Bhushan N. and Rai K., Strategic Decision Making Applying the Analytic Hierarchy Process, Springer, 2004[16] Bhushan N., Ideality, TRIZ and Software Design Case Study: Software Product for Identity Security,
Proceedings of TRIZCON 2008, US, 2008.
[17] Senge P.M., The Fifth Discipline, Random House, UK, 1990[18] Kim W. C. and Mauborgne R., Blue Ocean Strategy, Harvard Business Press, 2005[19] Gardner H, Frames Of Mind: The Theory Of Multiple Intelligences, Basic Books, 1983[20] Strogatz S., Sync: The Emerging Science of Spontaneous Order, Penguin, 2004[21] Gleick J, Chaos: The Amazing Science of the Unpredictable, Vintage, 1998[22] Fudenberg D.,Tirole J., Game Theory, MIT Press, 1991[23] Parsaei H.R., Sullivan W.G., Concurrent Engineering: Contemporary Issues and Modern Design Tools,
Springer, 1993
[24]
Forrester Jay. Industrial Dynamics, Pegasus Communications: Waltham, MA, 1961[25] Kugler S., Set-based concurrent engineering in Open Source Software Development,
www.vizzzion.org/stuff/thesis-final.pdf(accessed 02 October 2006)
[26] Dahan E., Reducing Technical Uncertainty in Product and Process Development through Parallel Design ofPrototypes, Graduate School of Business, Stanford University, 1998
[27] Wu S., A Probabilistic Model of Set-Based Design, MIT Undergraduate Journal of Mathematics[28] Pardes C. el al, Set-Based Design: A Decision-Theoretic Perspective,
http://www.pslm.gatech.edu/events/frontiers2006/proceedings/2006-03-16-Frontiers2006-
http://www.calresco.org/offline.htmhttp://www.vizzzion.org/stuff/thesis-final.pdfhttp://www.pslm.gatech.edu/events/frontiers2006/proceedings/2006-03-16-Frontiers2006-Paredis.pdfhttp://www.pslm.gatech.edu/events/frontiers2006/proceedings/2006-03-16-Frontiers2006-Paredis.pdfhttp://www.vizzzion.org/stuff/thesis-final.pdfhttp://www.calresco.org/offline.htm -
8/14/2019 Ultra Large Scale Systems Design Using LIST
16/16
Paredis.pdf#search=%22Set-Based%20Design%3AA%20Decision-Theoretic%20Perspective%22 (accessed on
02 Oct 2006)
[29] Rekuc, S.J., J.M. Aughenbaugh, M. Bruns, and C.J.J. Paredis, 2006, Eliminating Design Alternatives Based onImprecise Information, Society of Automotive Engineering World Congress, paper no. 2006-01-0272, April 3-6,
2006, Detroit, Michigan.
[30] Sobek D.K. et al., Toyotas principles of set-based concurrent engineering, Sloan Management Review, Vol.40, winter 1999.
[31] Ballard C., Positive Vs Negative Iteration in Design, Lean Construction Institution[32] Darcy D.P. et al, The Structural Complexity of Software: Testing the interaction of Coupling and Cohesion,
http://littlehurt.gsia.cmu.edu/gsiadoc/WP/2005-E23.pdf, accessed on 15th March 2005.
[33] Bonacich, P.B., Power and Centrality: A Family of Measures, American Journal of Sociology92, 1170-1182,1987
[34] Businessweek Research Services, Global Product Development Moving from Strategy to Execution, 2006.(Can be downloaded from http://www.PTC.com)
[35] Chavez T and Dion G.R., Using Decision Engineering to Achieve Short Predictable Lead Time at SunMicrosystems, Inc., Cycle Time Research, Vol. 4, No. 1, 1998.
[36] Dahan E., Reducing Technical Uncertainty in Product and Process Development through Parallel Design ofPrototypes, Graduate School of Business, Stanford University, 1998
[37] Eppinger S.D. and Tripathy A., DSM Models of Globally Distributed Product Development Structures, 8thInternational DSM Conference, Seattle, US, Oct 2006.
[38] Forrester Jay. Industrial Dynamics, Pegasus Communications: Waltham, MA, 1961[39] Hoch S.J., Kunreuther H.C. and Gunther R.E. (Eds), Wharton on Making Decisions, John Wiley & Sons, USA,
2006
[40] Jaiswal N.K., Military Operations Research Quantitative Decision Making, Kluwer Academic Publishers,Dordrecht, The Netherlands, 1997
[41] Larry H. and Sakkab N., Connect and Develop Inside Proctor and Gambles New Model of Innovation, HarvardBusiness Review, March 2006.
[42] Makajic-Nikolic D., et all, Bullwhip Effect and supply chain Modelling and Analysis Using CPN Tools,www.daimi.au.dk/CPnets/workshop04/cpn/papers/makajic-nikolic_panic_vujosevic.pdf (accessed 02 Oct
2006)
[43] Miller G.A., The Magical Number Seven, Plus or Minus Two: Some Limits on Our Capacity for ProcessingInformation, The Psychological Review, vol. 63, 1956
[44] Naughton K and Sloan A., Comin Through!, Newsweek, March 12, 2007. [45] Saaty T.L., The Analytic Hierarchy Process, McGraw Hill, NY, 1980[46] http://www.aitriz.org (The Altshuller Institute for TRIZ studies, website)[47] http://www.triz-journal.com[48] http://www.dsmweb.org[49] Iyer K, Holistic Value FrameworkCreating Right Value Streams Using TRIZ and Other Concepts, TRIZ Journal,
Jan 2007,http://www.triz-journal.com/archives/2007/01/02/
[50] Iyer K, Vedic Inventive Principles,http://www.innovationtools.com/PDF/Vedic_Inventive_Principles.pdf[51] Syvantek D. J., DeShon R.P., Organizational Attractors: A Chaos Theory Explanation of Why Cultural Change
Efforts Often Fail, Public Administration Quarterly, Fall 1993, 17, 3, ABI/INFORM Global
[52] Wujastyk D., The Roots of Ayurveda: Selections from Sanskrit Medical Writings, Penguin Classics, 2003*****
http://littlehurt.gsia.cmu.edu/gsiadoc/WP/2005-E23.pdfhttp://www.ptc.com/http://www.daimi.au.dk/CPnets/workshop04/cpn/papers/makajic-nikolic_panic_vujosevic.pdfhttp://www.aitriz.org/http://www.triz-journal.com/http://www.dsmweb.org/http://www.triz-journal.com/archives/2007/01/02/http://www.triz-journal.com/archives/2007/01/02/http://www.triz-journal.com/archives/2007/01/02/http://www.innovationtools.com/PDF/Vedic_Inventive_Principles.pdfhttp://www.innovationtools.com/PDF/Vedic_Inventive_Principles.pdfhttp://www.innovationtools.com/PDF/Vedic_Inventive_Principles.pdfhttp://www.innovationtools.com/PDF/Vedic_Inventive_Principles.pdfhttp://www.triz-journal.com/archives/2007/01/02/http://www.dsmweb.org/http://www.triz-journal.com/http://www.aitriz.org/http://www.daimi.au.dk/CPnets/workshop04/cpn/papers/makajic-nikolic_panic_vujosevic.pdfhttp://www.ptc.com/http://littlehurt.gsia.cmu.edu/gsiadoc/WP/2005-E23.pdf