1988 Fiscal Year Final Research Report Summary
Research on Resource Management in Distributed Computer Systems
| Project/Area Number |
62550260
|
|---|
| Research Category |
Grant-in-Aid for General Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Research Field |
計算機工学
|
|---|
| Research Institution | The University of Electro-Communications |
|---|
Principal Investigator |
KAMEDA Hisao The University of Electro-Communications, 電気通信学部, 教授 (10011660)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
ONOZATO Yoshikuni The University of Electro-Communications, 電気通信学部, 助教授 (10143710)
|
|---|
| Project Period (FY) |
1987 – 1988
|
| Keywords | Distributed Computer System / Computer Network / Load Balancing / Response Time / Overall Optimum / Individual Optimum / 衛星通信アクセス制御方式 / 多元接続方式 |
|---|
| Research Abstract |
(1) Studies on balancing computing loads in distibuted computer systems. Tantawi and towsley obtained the solutions of optimal load balancing strategies whereby the mean job response time is minimized in distributed computer systems with single channel networks and with star network configurations in the case of single class jobs. We call this the overall potimal policy. We extend their results in many directions. We first extend to the case of multiple class jobs. the optimal solution is obtained for tree network hierarchies which are multilevel star configurations. We find out an optimal load balancing algorithm that is simpler and more effecient than tantawi and towsley's. Furthermore we study the individually optimal policy whereby the user of each job finds that the expected resopnse thme of his job cannot be improved by changing the processing host. We show the existence of the solution for such a policy. We study the effect of system parameters by means of analysis and numerical examination, and find out some seemingly counter-intuitive phenomena.
(2) Studies on effective utilization of communication channels. Due to the enlargement of both the number of users and the transmission speed in a communication network, the requirements which are solicited from the network reach close up to the operations margin of the system and therefore some stability problems srise. We invwstigate the stability of different network architectures and network elements such as aloha-type access protocols and routing procedures. We give some ideas about how to buid up a network model which can accept the analytical tools supplied by the catastrophe theory and explained the system behavior beyond a stable operating mode. Based on those works we propose several network control protocols such as movable boundary and no boundary random access procedures.
|
