Font Size: a A A

Study On Network Performance Measurement And Service Optimization In Cloud Computing

Posted on:2017-03-01Degree:DoctorType:Dissertation
Country:ChinaCandidate:H DingFull Text:PDF
GTID:1108330482972291Subject:Communication and Information System
Abstract/Summary:PDF Full Text Request
Cloud computing, which is based on the current Internet, has established a new structure for information and resource allocation in the form of cloud service. Greatly fulfilled its customers, it has also become the implementation of many Service Science concepts. With the increasing in the number of cloud services and customers, the cloud platform owner and clouds service providers now have to consider network environment, resource distribution issues and service patterns carefully in their business. They are facing challenges in terms of efficiency and benefit in their service.To address the issue of service optimization for cloud customers, we first classify cloud customers into two types based on the characteristic of their required services:lightweight cloud customer and heavyweight cloud customer. Lightweight cloud customers consume less service resources, occupy shorter service time and have higher real-time requirements, while heavyweight cloud customer are just the opposite:they consume more service resources, occupy relatively longer service time and almost do not have real-time requirements. This difference further classifies the service optimization issue for each type of customers:lightweight cloud customers are facing a service request optimization issue, while heavyweight cloud customers are facing a service deployment optimization issue. Based on these, we consider the cloud service optimization issue in this dissertation. By investigating the current network, we find problems of it, and then propose solutions for these two types of cloud customers separately. We further verify our solutions by conducting simulation experiments. The research efforts we get in this dissertation will have great meaning for cloud service providers in the industry.The main research efforts and innovative points are as follows:(1) We propose an innovative packet loss prediction method, and correct the misleading discriptions of the TCP timestamp option utilization rate in the acdemic community. Meanwhile, to our knowledge, we are the first to measure the actual usage of so-called "stetch ACKs" on the Internet. Moreover, through comparing experiments, we further analyze advantages and disadvantages of different RTT measurement technologies and the actual RTT distribution.(2) To address the problem found in point (1), we give an optimization structure for the whole could network. Under the structure, we propose an optimization strategy for lightweight cloud customers in terms of the service request issue. This strategy includes a cloud server workload prediction method and corresponding service request optimization algorithm. The prediction method and optimization algorithm are proved useful and effective by simulation experiments..(3) We further consider the service deployment issue for heavyweight cloud customers, model this issue into a modified quadratic assignment issue using set theory and matrix theory, and then use simulated annealing algorithm to optimize it. We give out all the details of the improved algorithm, and verify its usability and effectiveness by numerical simulations.
Keywords/Search Tags:network measurement, cloud service, service request, service deployment, service optimization
PDF Full Text Request
Related items