A Market for Scheduling, with Applications to Cloud Computing
Hausdorff Center for Mathematics via YouTube
Overview
Explore a groundbreaking market model for allocating and scheduling resources in cloud computing presented by Ruta Mehta. Delve into a novel approach that addresses two critical aspects: agents requiring specific resource amounts for task completion and the desire for swift task execution. Discover how this model diverges from traditional market models in economics and theoretical computer science, necessitating new techniques for computing equilibria. Learn about a polynomial-time algorithm that utilizes a method reminiscent of Myerson's ironing procedure for optimal auctions. Understand the complexities of equilibrium prices, including potential non-convexity and "holes" in the price set. Gain insights into the motivation behind this research, driven by the rapidly growing cloud computing marketplace and its current ad hoc management. Explore topics such as the scheduling problem, Amazon EC2, issues with the Spot Market, equilibrium characterization, optimal bundle LP, and generalizations for weighted flow time and multiple resources.
Syllabus
Intro
Cloud computing market
Scheduling problems
Amazon EC2
Issues with Spot Market
Model
Equilibrium
Characterization
Optimal Bundle LP
Single Good
Algorithm
Arbitrary Goods
Weighted Flow Time
Computing Prices
Generalization
Resources
Taught by
Hausdorff Center for Mathematics