On Minimizing Cost Of Datacenters In Hybrid Cloud

With development of cloud computing,datacenters are becoming more and more popular.Datacenters can provide scalable resources for applications such as big data processing,service computing,which also bring high operation costs.Thus,it is critical to minimize cost of datacenters.In this paper,we study three problems on minimizing the cost of datacenters.We develop approximation algorithms and theoretical analyze the performance of algorithms.Firstly,we study VM scheduling problem to minimize time-variable electricity cost in cloud systems.We investigate consolidation strategies by incorporating electricity price variations to reduce its electricity bill.We first develop an offline algorithm with an approximation ratio of 5,which always approaches the optimal solution within 5 times.For the online setting,we propose a 0(Lmax/Lmin)-competitive algorithm that is able to approach the optimal offline solution within a factor of 0(Lmax/Lmin)in the worst case,where Lmax/Lmin is the ratio of the longest length of the process time requirement of the VMS to the shortest one.Our trace-driven simulation results further demonstrate that the average performance of the proposed algorithms can lead to near optimal solution.Then,we study task scheduling problem to minimize service provisioning cost of hybrid clouds with flash intensive requests.We leverage hybrid cloud solution by renting services from the public clouds and offloading the excessive user request to public clouds.We design an online optimization algorithm for minimizing the service provisioning cost in hybrid cloud based on Lyapunov optimization.The algorithm achieves the service provisioning cost that is arbitrarily closed to the optimal solution.Simulation results show that our algorithm can efficiently reduce the service provision cost for hybrid clouds facing flash intensive requests.Finally,we study the scheduling problem that minimizes service provisioning cost of hybrid clouds with flash intensive requests and request dropping.We notice that in practice,the requests can be partially dropped under intensive cases and served through re-requesting.Hybrid cloud can choose some requests to drop so as to avoid unlimited growth of waiting queue and ensure the worst delay of requests processing.To minimize the service provisioning cost in hybrid cloud,we design an online algorithm based on Lyapunov optimization.The algorithm achieves the service provisioning cost arbitrarily closed to the optimal solution.Simulation results show that our algorithm can efficiently reduce the service provision cost for hybrid clouds facing flash intensive requests.