EM.411 meeting
Bruce Cameron
Making decisions requires value curves
Value curve
Tradespaces
Elicit trade off between two variables
Pareto front
The line drawn on a two variable tradespace between all the non dominated options
Every dot is a valid feasible design
Daniel Kahneman
Our intuition may be worse on important problems; certainly no reason to believe they are positively correlated
Architectural Decisions
the most impactful decisions that produce different architectures
tend to be mutually exclusive
controllable
ie my idea of a design choice
design choice
a decision within the control of the creator of a system
then Architectural Decisions are a subset of design choice
upstream of metrics
e.g. can’t dictate a “reliability” metric
Exercise on a system
Glider architectural decisions
Engine
Passenger capacity
Wing type
Wing position
Landing gear configuration
Material
tests for architectural decisions
sensitivity
changing x significantly changes system outcomes
connectivity
harder to change later because there is a lot of interdependence
Is connectivity actually different than cost ?
Good architectural decisions are like finding the end of the ball of yarn
it makes a lot of decisions with one decision
High sensitivity and high connectivity decisions should be made earlier and with more care