Dynamic System Optimization through Performance Modeling Universality and

Dynamic System Optimization through Performance Modeling: Universality and Decomposition Y. C. Tay National University of Singapore

Dynamic optimization is hard. Examples: (1) Prob(miss) = f(cache size) (2) Given target Prob(miss), how to dynamically adjust cache size? Universality (2) Internet traffic equilibrium How to prevent performance collapse from congestion? Decomposition

(1) Universality issue: Prob(miss) = f(cache size) complex reference cache management pattern size policy add: ● ● data layout hardware variation application mix data instance system configuration software variation system customization hw/sw evolution autonomic configuration dynamic adjustment intractable!

(1) Universality no change idea: Prob(miss) = f(cache size | parameters) change values ● ● system customization hw/sw evolution autonomic configuration dynamic adjustment

(1) Universality idea: Prob(miss) = f(cache size | parameters) #miss = f (M | M*, M 0, n*, n 0) = 1 2 change values (H + √(H 2 – 4))(n*+n 0) – n 0 where H ● ● = 1+ M* - M 0 M - M 0 system customization hw/sw evolution autonomic configuration dynamic adjustment

(1) Universality #miss = f (M | M*, M 0, n*, n 0) = Example: cache=RAM 1 2 (H + √(H 2 – 4))(n*+n 0) – n 0 where H = 1+ M* - M 0 M - M 0