1990 Fiscal Year Final Research Report Summary
Research on the Design of Distributed Operating Systems
| Project/Area Number |
01460255
|
|---|
| Research Category |
Grant-in-Aid for General Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Research Field |
Informatics
|
|---|
| Research Institution | The University of Electro-Communications |
|---|
Principal Investigator |
KAMEDA Hisao Univ. Electro-Comm., Fac. E-C., Professor, 電気通信学部, 教授 (10011660)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
砂原 秀樹 電気通信大学, 電気通信学部, 助手 (20206577)
KAKUDA Hiroyasu Univ. Electro-Comm., Fac. E-C., Assist. Prof., 電気通信学部, 講師 (70152600)
SHIMIZU Kentaro Univ. Electro-Comm., Fac. E-C., Assoc. Prof., 電気通信学部, 助教授 (80178970)
TERASHIMA Motoaki Univ. Electro-Comm., Fac. E-C., Assoc. Prof., 電気通信学部, 助教授 (20114928)
KASAI Takumi Univ. Electro-Comm., Fac. E-C., Professor, 電気通信学部, 教授 (70027382)
|
|---|
| Project Period (FY) |
1989 – 1990
|
| Keywords | Distributed Operating System / Distributed Computer System / Optimal Load Balancing / Load Balancing Algorithm / Parametric Analysis / Process Migration / Light-Weight Process / Computer Network |
|---|
| Research Abstract |
A distributed computer system consists of a number of host computers connected by a communications network. The operating system which makes a distributed computer system to operate as if it were an integrated system is called a distributed operating system. We obtained the following results on distributed operating systems.
1. Distributed operating systems balance the loads on the host computers so that the host computers may be used efficiently. We studied optimal policies for such load balancing. First, we provided an improved version of the Tantawi and Towsley single job class load balancing algorithm, and extended it to a multiclass version. We showed that these proposed algorithms have better performance with respect to the computation time and memory requirements than the previous ones. Then we analyzed the effects of system parameters on the optimal load balancing policies analytically in the case of the single job class and experimentally in the case of the multiple job classes. In both cases, we found various intriguing properties of the optimal load policies. Furthermore, we obtained a decomposition algorithm for optimal load balancing in tree hierarchy network configurations.
2. We studied the literature in this field and recognized the importance of page turning policies during process migration and policies for managing distributed shared memory. We designed a simulator that can be used to evaluate the performance of network paging.
3. We studied process structuring which forms the basis of distributed operating systems. We performed the conceptual design of "micro-processes" which are implemented mostly in the out-of-kernel level. Furthermore, we implemented light-weight processes in the out-of-kernel level of the UNIX operating systems, evaluated their performance and examined the underlying problems.
|
