2004 Fiscal Year Final Research Report Summary
Design and Analysis of Scalable Web Servers
| Project/Area Number |
14380139
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
計算機科学
|
|---|
| Research Institution | Japan Advanced Institute of Science and Technology |
|---|
Principal Investigator |
SHEN Hong Japan Advanced Institute of Science & Technology, School of Information Science, Professor, 情報科学研究科, 教授 (60333556)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
HORIGUCHI Susumu Tohoku University, Graduate School of Information Science, Professor, 大学院・情報科学研究科, 教授 (60143012)
JIANG Xiaohong Tohoku University, Graduate School of Information Science, Associate Professor, 大学院・情報科学研究科, 助教授 (00345654)
|
|---|
| Project Period (FY) |
2002 – 2004
|
| Keywords | Internet Computing / Web Technology / Object Caching / Content Replacement / Replication / Cache Placement / Proxy Placement / Replication |
|---|
| Research Abstract |
In this project we have studied new methods and techniques to design scalable Web servers through Web caching and service replication. We have integrated these techniques together to build up a prototype Web server that is scalable at different levels. Our project has two streams followed by an integration and prototyping phase, where all the tasks are implemented and tested, with an extensive performance comparison with existing results.
1.Caching : Caching effectively migrates copies of popular documents from Web servers closer to the Web clients, and therefore reduces response time and network traffic. In the first project year we studied how large caches should be installed at client servers and proxy servers respectively so that the cost-effectiveness is optimized. In the second project year we applied various techniques to discover network topologies and access frequencies at each client server.
2.Replication : Replication is another form of data migration in order to improve syste
… More
m performance. It copies and stores data at dedicated servers (mirror servers), instead of in local caches as described in the previous stream, which are geographically close to the client.
In the final project year concentrated on tasks of deciding the contents for replacement when local caches are full for the first stream, and developing effective models to keep consistency among replicas for the second stream.
In addition to completion of the above proposed tasks in the project, we have also completed the following tasks related to the project theme : We have proposed a novel approach to analyzing the performance of optical switching networks and a new scheme to realize crosstalk-free permutation in vertically stacked optical MINs. We have studied the crosstalk-free permutation in photonic rearrangeable networks built on a combination of horizontal expansion and vertical stacking of banyan networks.
We have developed effective algorithms on communication patterns (topology) discovery in the Internet. Some important analysis and comparison between previous methods and our proposed new methods have been done. We have also proposed a mobile agents based model for network topology discovery.
We have successfully designed a highly scalable wavelet-based image codex, S-SPECK, which achieves distortion scalability, resolution scalability, and ROI retrievability in an integrated framework. We have proposed a novel robust blind wavelet-based image watermarking algorithm to make watermarking invulnerable to geometric attacks.
We have modified conventional SVM, Robust SVM and One-class SVM techniques for host-based intrusion detection respectively based on the idea from Online SVM, and their performance were compared with that of the original ones. Both the theoretical analysis and experiments indicate that our methods can be trained online and outperform the original ones with fewer SVs and less training time without decreasing detection accuracy. Less
|
Research Products
(128 results)
