Ultra Large Scale Systems Design Using List

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. Introduction “The 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

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, nonlinearity, 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 as a 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 were based 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

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 Needs Given the systems that we have built and which are continuing to scale-up in all walks of life, we are closer 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-physicalsocial 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, and changing 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 Decentralized Control

Present Approaches All conflicts must be resolved and resolved centrally and uniformly conflicting, Requirements can be known in advance and change slowly. and diverse Tradeoff decisions will be stable.

Inherently unknowable, requirements Continuous evolution and System improvements are introduced at discrete intervals. deployment Heterogeneous, inconsistent, Effect of a change can be predicted sufficiently well. and changing elements Configuration information is accurate and can be tightly controlled. Components and users are fairly homogeneous. Erosion of the People are just users of the system. Collective behavior of people people/system boundary is not of interest. Social interactions are not relevant. Normal Failures Failures will occur infrequently. Defects can be removed. New paradigms for A prime contractor is responsible for system development, acquisition and policy 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

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 – in fact nature’s 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 Human Interaction

Computational Emergence

Design

Computational Engineering

Adaptive System Infrastructure

Specific Sub-Areas • Context-Aware Assistive computing • Understanding Users and Their Contexts • Modeling Users and User Communities • Fostering Non-Competitive Social Collaboration • Longevity • Algorithmic Mechanism Design • Metaheuristics in Software Engineering • Digital Evolution • Design of All Levels • Design Spaces and Design rules • Harnessing Economics to Promote Good Design • Design Representation and Analysis • Assimilation • Determining and Managing Requirements • Expressive Representation Languages • Scaled-Up Specification, Verification, and Certification • Computational Engineering for Analysis and Design • Decentralized Production Management • View-Based Evolution • Evolutionary Configuration and Deployment • In Situ Control and Adaptation

Adaptable and Predictable System Quality

Policy, Acquisition, and Management

• 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 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 Systems Approach 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.

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.Needs What 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 intrapersonal. 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 beneficially impact 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

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 multicontextual and evolving map of needs. A multilevel user needs dashboard can serve as a key tool in designing the 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

(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 of resources 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 one can 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 points where 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 system evolution. 3. Choose a design which is moving towards ideality – all benefits at zero cost.

3.3.Structure Often, 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?

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 cannot afford to fail b. Unpredictability – They are not very well understood and are unpredictable, hence they are deliberately operating within strict constraints c. Dependency – Too many other system elements are dependent on this element, hence it cannot be changed very easily and without pain d. Insulation – They are not very well connected and therefore do not have an incentive to change or adapt to changes e. Efficiency – The elements are optimally structured to perform certain functions as efficiently as possible All of these reasons are fundamentally linked to the concept of system complexity as defined by couplingcohesion 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) using Dependency 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.Behavior While 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.

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 USA’s 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 clocks can 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 to know 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.

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 will get 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 that the 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. 2. 3. 4.

Fire – fuel, energy to run the system Water – vitality or life of the system, ideas Air – transports essentials to all parts of the system, enables change Earth – structure and raw material, the basis of the system

In 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

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-based Approach 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 Optima As 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 complexity The 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 delay Being forced to choose the best alternative may result in procrastination – “Let 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

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 Engineering Set-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 shorter development 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 Toyota’s 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 been termed 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 don’t 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

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 to strengths of different teams be identified  Let each team explore the specifications, needs, functions independent of each other  Each team explore design tradeoffs through simulations and their past observations  Each team should come up with their sets of different solutions with in the functional and performance needs of the product

 Problem Formulation and Analysis  Value Stream [49]  Ideal Final Result (IFR)  Why-what hierarchy  Nine windows  Dependency Structure Matrix (DSM) [48]  Function/Attribute Analysis  System Complexity Estimator (SCE) [12][16]  S curve analysis  Vedic Inventive Principles [50]  Contradictions – Technical/Physical  Trends of evolution

Striving for  Design should remain functional after variations Conceptual in its environment Robustness  Vulnerability of system to changes in the (Functional Team environment should be minimized level)  Modularized Design with standard components

 IFR  AFD/Subversion Analysis  Robust Inventive System Design (RISD) [10]  DSM

 Decision Dependency Matrices (DDM) [13][15]  Analytic Hierarchy Process (AHP) [15][45]  Technical Contradictions / Inventive Principles Establish  Multiple concepts developed using prototyping  Decision theoretic Feasibility before simulation principles [28][29] Commitment  The infeasible ones will be rejected rest all will  AHP continue to be developed  Closer to IFR Conflict Handling Cooperative Conflict handling  Which solution is closer to IFR?  DDM  AHP Integration Intersection (System level)

by  How are the parts integrated to meet at the point that will be regarded best solution  Find out overlap of feasible design spaces for each sub component  Decisions about eliminating the weak designs

6. Conclusions and Further Research The 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

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] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12]

[13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28]

Anderson, P.W., More is Different, Science, Vol. 177, No. 4047, Aug. 4, 1972, pp. 393-396. Northrop L., Ultra-Large-Scale Systems – The Software Challenge of the Future, SEI Report, June 2006. Gershenson C. and Heylighen F., How Can We Think the Complex?, http://www.calresco.org/offline.htm (accessed on 12 August 2006) Ray Kurzweil, The Singularity is Near –When Humans Transcend Biology, Penguin, US, 2005. Simon H. A., The Sciences of the Artificial. 3rd edition, The MIT Press, USA, 1996. Northrop L., Scale Changes Everything, ICGSE, Brazil, 2006. May Mathew E., The Elegant Solution – Toyota’s Formula for Mastering Innovation, Free Press, NY, 2007 Liker J.K, The Toyota Way: 14 Management Principles from the World’s Greatest Manufacturer, McGraw-Hill Professional, 2004 Bhushan N., Set-Based Concurrent Engineering (SBCE) and TRIZ – A Framework for Global Product Development, Proceedings of TRIZCON 2007, Altshuller Institute, 2007 Bhushan N, Robust Inventive Software Design – A Framework Combining DSM, AHP and TRIZ, 7th DSM Conference, Seattle, US, 2005. Bhushan N. and Iyer K., Strategic Project Management Priorities in Global Software Development Scenarios, Communicated to IEEE Software, 2007. 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. Bhushan N., Decision Dependency Matrices, 8th International DSM Conference, Seattle, US, Oct 2006 Bhushan N., Set Based Concurrent Engineering Using DSM – A Framework for Software Development, 8th International DSM Conference, Seattle, US, Oct 2006. Bhushan N. and Rai K., Strategic Decision Making – Applying the Analytic Hierarchy Process, Springer, 2004 Bhushan N., Ideality, TRIZ and Software Design – Case Study: Software Product for Identity Security, Proceedings of TRIZCON 2008, US, 2008. Senge P.M., The Fifth Discipline, Random House, UK, 1990 Kim W. C. and Mauborgne R., Blue Ocean Strategy, Harvard Business Press, 2005 Gardner H, Frames Of Mind: The Theory Of Multiple Intelligences, Basic Books, 1983 Strogatz S., Sync: The Emerging Science of Spontaneous Order, Penguin, 2004 Gleick J, Chaos: The Amazing Science of the Unpredictable, Vintage, 1998 Fudenberg D.,Tirole J., Game Theory, MIT Press, 1991 Parsaei H.R., Sullivan W.G., Concurrent Engineering: Contemporary Issues and Modern Design Tools, Springer, 1993 Forrester Jay. Industrial Dynamics, Pegasus Communications: Waltham, MA, 1961 Kugler S., Set-based concurrent engineering in Open Source Software Development, www.vizzzion.org/stuff/thesis-final.pdf (accessed 02 October 2006) Dahan E., Reducing Technical Uncertainty in Product and Process Development through Parallel Design of Prototypes, Graduate School of Business, Stanford University, 1998 Wu S., A Probabilistic Model of Set-Based Design, MIT Undergraduate Journal of Mathematics Pardes C. el al, Set-Based Design: A Decision-Theoretic Perspective, http://www.pslm.gatech.edu/events/frontiers2006/proceedings/2006-03-16-Frontiers2006-

[29]

[30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42]

[43] [44] [45] [46] [47] [48] [49] [50] [51] [52]

Paredis.pdf#search=%22Set-Based%20Design%3AA%20Decision-Theoretic%20Perspective%22 (accessed on 02 Oct 2006) Rekuc, S.J., J.M. Aughenbaugh, M. Bruns, and C.J.J. Paredis, 2006, Eliminating Design Alternatives Based on Imprecise Information, Society of Automotive Engineering World Congress, paper no. 2006-01-0272, April 3-6, 2006, Detroit, Michigan. Sobek D.K. et al., Toyota’s principles of set-based concurrent engineering, Sloan Management Review, Vol. 40, winter 1999. Ballard C., Positive Vs Negative Iteration in Design, Lean Construction Institution 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. Bonacich, P.B., Power and Centrality: A Family of Measures, American Journal of Sociology 92, 1170-1182, 1987 Businessweek Research Services, Global Product Development – Moving from Strategy to Execution, 2006. (Can be downloaded from http://www.PTC.com) Chavez T and Dion G.R., Using Decision Engineering to Achieve Short Predictable Lead Time at Sun Microsystems, Inc., Cycle Time Research, Vol. 4, No. 1, 1998. Dahan E., Reducing Technical Uncertainty in Product and Process Development through Parallel Design of Prototypes, Graduate School of Business, Stanford University, 1998 Eppinger S.D. and Tripathy A., DSM Models of Globally Distributed Product Development Structures, 8th International DSM Conference, Seattle, US, Oct 2006. Forrester Jay. Industrial Dynamics, Pegasus Communications: Waltham, MA, 1961 Hoch S.J., Kunreuther H.C. and Gunther R.E. (Eds), Wharton on Making Decisions, John Wiley & Sons, USA, 2006 Jaiswal N.K., Military Operations Research – Quantitative Decision Making, Kluwer Academic Publishers, Dordrecht, The Netherlands, 1997 Larry H. and Sakkab N., Connect and Develop Inside Proctor and Gamble’s New Model of Innovation, Harvard Business Review, March 2006. 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) Miller G.A., The Magical Number Seven, Plus or Minus Two: Some Limits on Our Capacity for Processing Information, The Psychological Review, vol. 63, 1956 Naughton K and Sloan A., Comin’ Through!, Newsweek, March 12, 2007. Saaty T.L., The Analytic Hierarchy Process, McGraw Hill, NY, 1980 http://www.aitriz.org (The Altshuller Institute for TRIZ studies, website) http://www.triz-journal.com http://www.dsmweb.org Iyer K, Holistic Value Framework – Creating Right Value Streams Using TRIZ and Other Concepts, TRIZ Journal, Jan 2007, http://www.triz-journal.com/archives/2007/01/02/ Iyer K, Vedic Inventive Principles, http://www.innovationtools.com/PDF/Vedic_Inventive_Principles.pdf 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 Wujastyk D., The Roots of Ayurveda: Selections from Sanskrit Medical Writings, Penguin Classics, 2003 *****