Capacity planning for computing clusters

This disclosure provides techniques that enable elastic computing service providers to optimize allocation of computing resources across multiple tiers of service level objectives, while allowing for periods of oversubscription to be met by cross-tier movement of resources. Further, the techniques enable providers of computing services to plan their computing investments given rates of oversubscription and anticipated demand across tiers. Service providers frequently plan investments based on peak, rather than average, demand and may therefore be left with surplus capacity during periods of below-peak demand. While such surplus can be resold at lower levels of service guarantees, such reselling increases demand (oversubscription). By accounting for oversubscriptions and inflations of service level objectives across tiers, the techniques of this disclosure enable optimal multi-tier resource allocation