Copyright 2005 Northrop Grumman Corporation 0 Critical Success Factors for system-of-system architecture / engineering 25 October 2006 Neil Siegel Sector Vice-President, Technology Northrop Grumman Mission Systems

Copyright 2005 Northrop Grumman Corporation 1 SoS design basics – one person’s view  Understand the work-flow and dynamic behavior of the system  “Long mission threads”  Simultaneous operations  Capacity  Port-to-port timing  Etc.  Separate the implementation of the structure from the implementation of the “meat”  Exercise the structure at scale early and often  Design for allowed / unallowed dynamic behavior  “DC timing diagram” analogy  Establishing boundaries and linkages  Tight versus loose coupling  Where to use each A small number of abstractions (“views”) are helpful. Science + engineering + art. Simplicity is a virtue.

Copyright 2005 Northrop Grumman Corporation 2 Fitting the work to the distribution of skills in a real team  Implementation can vary significantly in complexity  In any large team, skill level across the team will vary significantly  Matching significantly improves the likelihood of a desirable outcome  Specific, tangible design steps to partition the design into “zones of pre-determined implementation complexity”

Copyright 2005 Northrop Grumman Corporation 3 Design for quality factors  We are better as an industry at designing for highly- visible factors (e.g., functionality) than for underlying quality attributes (e.g., MTBF)  Real-world practice seems to result in a huge variance in achieved quality  Probably an indication of an immature state-of-practice across our industry  Not clear if the underlying cause is lack of skill or lack of focus  A big killer of programs!

Copyright 2005 Northrop Grumman Corporation 4 Accept the limitations of our intuition  The relationship between improvement in a technical factor and improvement in observed system performance is not always obvious  Yet we all too often depend on intuition, or make use of hidden assumptions that do the same thing  Link technical predictors to operational predictors

Copyright 2005 Northrop Grumman Corporation 5 Integrating people into the system’s business process  Building successful system-of-systems is a process that must include a business-process- reengineering aspect  Understand the sociology of your user community, not just what they say  Perform a careful partitioning of what the human can do best, and what the computer can do best  Effective and credible  Don’t ask the human for data the computer can figure out  Support the stressed user  Crossing security domains

Copyright 2005 Northrop Grumman Corporation 6 Summary  At system-of-systems scale, things sometimes scale badly / non-intuitively  Unplanned dynamic behavior is the source of many of the hard problems  Not all people are equally skilled, so designs that are based on a tacit assumption that they are all equally skilled are risky  The large dynamic range of quality outcomes is a sign of immature state-of-practice Informing the system-of-systems design through domain knowledge seems essential