1997 Fiscal Year Final Research Report Summary
A Study on Scheduling Disciplines for Providing the Quality of Services Guarantee
| Project/Area Number |
08680446
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
情報システム学(含情報図書館学)
|
|---|
| Research Institution | National Center for Science Information Systems |
|---|
Principal Investigator |
JI Yusheng National Center for Science Information Systems, Research and Development Dept., Associate Professor, 研究開発部, 助教授 (80225333)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
ZHAO Weiping National Center for Science Information Systems, Research and Development Dept.,, 研究開発部, 助手 (80260016)
|
|---|
| Project Period (FY) |
1996 – 1997
|
| Keywords | scheduling / service discipline / quality of services / resource reservation / fairness |
|---|
| Research Abstract |
This research project is to study on the scheduling disciplines which can guarantee the quality of services by reserving system resources to individual applications in multimedia information systems connected by high speed communication networks.
For this purpose, we have studied and evaluated several scheduling disciplines. There are two main kind of scheduling desciplines ; One is the first-come-first-served scheme, which shares buffer among different logic associations. The other is based on the Processor Sharing scheme, in which objects are served in turn by assigning different buffers for different associations. Through our study, we found that disciplines with different buffers for different associations are feasible to reserve resources and hence to guarantee the quality of services. We also proposed the implementation method of such scheduling discipline by extending the Viertual Clock notation.
The fairness criterion, which concerns about how to allocate resource among applications, is an important issue when considering scheduling disciplines. Criteria established by other studies are all trying to provide a fair share among applications in every instant of time. On the other hand, with service requests which arrive in bursts like lots of real time applications, there needs a more flexible scheduling discipline to provide higher responsibility. Towards this end, we defined a more general fairness criterion, which is expected to be more suitable for bursty service requests. Under this new criterion, a class of scheduling discipline can be produced. We evaluated their performance and effects on various traffic analytically, and by simulations. We found it can be expected having more desirable features.
|
Research Products
(18 results)
