|
The problem on hand consists of optimizing a set of objectives, some of which
might be conflicting with one another. For example, achieving better latency
might require creating additional replicas. However, doing so would invariably
increase storage and consistency maintenance costs. Similarly, placing the PC network configuration replicas
on most stable nodes might not be ideal from latency minimization perspective.
Multi-objective optimization deals with these conflicting objectives by
evolving a set of solutions that compromise these conflicting objectives.
The quality of solutions obtained by multi-objective optimization is inherently
dependent upon how well the objectives are formulated. In this section, we first
model each of the objectives following which we discuss their conflicts.
|
