Fast Optimization for Scalable Application Deployments in Large Service Centers.

详细信息   

• 作者：Li ; Jim Zhanwen.
• 学历：Doctor
• 年：2011
• 毕业院校：Carleton University
• ISBN：9780494815830
• CBH：NR81583
• Country：Canada
• 语种：English
• FileSize：6763200
• Pages：169

文摘

Large complex service centers such as clouds must provide many services to many users with different service contracts. To deploy applications with many tasks across a cloud infrastructure, many goals must be satisfied simultaneously, which poses a large and complex optimization problem. The goals include quality-of-service targets, power minimization, respecting limited memory and software licences, and ability to track changing demands. This thesis creates new approaches to task assignment in service centers that are QoS driven and fast, scalable and extensible to new objectives and classes of restrictions. The new approaches provide sound configurations across a large number of highly coupled decisions. These approaches can effectively handle online resource management in various difficult dynamic environments, increasing the stability of management. The new approaches include i) the combination of linear optimization with an analytical performance model to solve a complex nonlinear optimization problem via a series of LP solutions, ii) a mixed-integer linear programming MIP) model to address many interacting integer constraints, iii) a new heuristic to approximately solve large-scale MIPs with greater efficiency, and iv) strategies to increase robustness in the face of dynamic changes in demand.