nA
way to think about system design
oCould I do this lazily/speculatively?
oWhen would it pay?
nSteps
toward a sound theory of laziness or speculation
oI am not presenting such a theory
nLazy:
defer work that may not be needed
oPays in saved work (and perhaps in latency)
oPays in more concurrency (if you have extra
resources)
nSpeculative:
Do work that may not be needed
oPays in more concurrency (if you have extra resources)