Optimal Web-scale Tiering as a Flow Problem
G Leung, Novi Quadrianto, AJ Smola and K Tsioutsiouliklis
In: Advances in Neural Information Processing Systems (NIPS)(2011).
We present a fast online solver for large scale parametric max-ﬂow problems as they occur in portfolio optimization, inventory management, computer vision, and logistics. Our algorithm solves an integer linear program in an online fashion. It exploits total unimodularity of the constraint matrix and a Lagrangian relaxation to solve the problem as a convex online game. The algorithm generates approximate solutions of max-ﬂow problems by performing stochastic gradient descent on a set of ﬂows. We apply the algorithm to optimize tier arrangement of over 84 million web pages on a layered set of caches to serve an incoming query stream optimal.