| Author: | Lai, Kunfeng |
| Title: | Towards migrating user applications to the cloud |
| Degree: | Ph.D. |
| Year: | 2012 |
| Subject: | Cloud computing. Systems migration. Hong Kong Polytechnic University -- Dissertations |
| Department: | Department of Computing |
| Pages: | xiv, 143 p. : ill. ; 30 cm. |
| Language: | English |
| Abstract: | The cloud computing has become an increasingly popular paradigm for Internet service hosting. It increases the operational efficiency by freeing the users from the trivial matters, such as software and hardware con.guration, electricity power management, to name but a few. To run their applications on the cloud, enterprises should migrate their data, computing, and operations to the cloud side. Though this migration trend is clear, there are many difficulties. The applications migrated to the clouds usually face multiple online users with widely diverse requirements; and the number of users can sharply increase or decrease. Therefore, it is challenging for an enterprise to migrate its applications to the clouds while providing stable services to the users with minimum infrastructure costs and minimum resource consumption. In this thesis, we conduct a systematic study on this problem. First, we conduct studies on user traffic pattern as we believe the migration is tightly coupled with user behavior. Second, we design a stable workload management scheme that maintains service quality and adapts to user traffic dynamics. Third, we study how to deploy applications to the cloud servers with minimized deployment cost. Our thesis completes the cycle with user behavior understanding, user service quality maintenance and cost minimization. Our results show that the user traffic of the Internet applications can be affected by various factors. For example, the external links may increase the video popularity in the video sharing site like Youku by 15%. Our workload management scheme can stably optimize the resource allocation under the dynamics of user traffic, and our deployment scheme can save the costs by 30% as compared to the traditional deployment scheme. In detail, we firstly study how one special factor of the video sharing sites, external links, affects the video sharing sites popularity. These external links is provided by video sharing sites, for videos to be embedded into external web pages. Clearly, the purpose of such function is to increase the distribution of the videos. Does this function fulfill its purpose? In this part of our thesis, we provide a comprehensive measurement study and analysis on these external links to answer this question. With the traces collected from two major video sharing sites, YouTube and Youku of China, we show that the external links have impacts on the popularity of the video sharing sites. More specifically, for videos that have been uploaded for eight months in Youku, around 15% of views can come from external links. We also show that if a video is popular itself, it is likely to have a large number of external links. In the second scenario, we design a stable workload management scheme to adapt for user traffic dynamics. To assure the service quality, the Service Level Agreement (SLA) is widely adopted. However, as the resource of the service provider is limited and the user requests are dynamic and online, the SLA can be violated. This will incur a loss, in money or in the number of users. As such, an important objective of the service provider is to minimize such loss. A widely adopted scheme for SLA maintenance is based on the prediction of the request arrival rate. In the study of this part, we will show that such scheme cannot always achieve a good minimization as they neglect the close loop between the predicted request arrival rate and response time. We propose an improved workload management scheme based on Iterative Extended Kalman Filter that can optimize and stabilize the system under SLA constraints. Based on analysis and real experiments, we show that our scheme can minimize the loss from SLA violation and achieve stability under fluctuating request arrival rates. Finally, we further consider the deployment schemes with the objective of minimization the deployment cost. To deploy the Internet applications, nowadays, end users purchase from one service provider. As there are an increasing number of platforms that can support applications to run cross different providers, we propose a multi-provider scheme to select different service instances from multiple providers to save costs in this part of our thesis. We formulate this multi-provider selection problem. We show that the problem is NP-complete. We develop an optimal algorithm for a special case and a non-trivial (1+ ε) approximation algorithm for the general problem. We further study two variants and develop efficient algorithms. We systematically evaluate our algorithms under different configurations and we conduct a case study for a real online MMORPG web-game. We show that by sharing services on Amazon, GoGrid and Rackspace, we can reduce the bill by 30%. |
| Rights: | All rights reserved |
| Access: | open access |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| b26158747.pdf | For All Users | 3.75 MB | Adobe PDF | View/Open |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/6976