Load Diffosion Style Network Soper Compvting
| Project/Area Number |
10480059
|
|---|
| 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 | KYOTO UNIVERSITY |
|---|
Principal Investigator |
TOMITA Shingi Kyoto University, Gradvate School of Informatics, Professor, 情報学研究科, 教授 (40026323)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
GOSHIMA Masahiro Kyoto University, Gradvate School of Informatics, Research Associate, 情報学研究科, 助手 (90283639)
MORI Shin-ichiro Kyoto University, Gradvate School of Informatics, Associate Profes, 情報学研究科, 助教授 (20243058)
|
|---|
| Project Period (FY) |
1998 – 1999
|
| Project Status |
Completed (Fiscal Year 1999)
|
| Budget Amount *help |
¥12,800,000 (Direct Cost: ¥12,800,000)
Fiscal Year 1999: ¥6,700,000 (Direct Cost: ¥6,700,000)
Fiscal Year 1998: ¥6,100,000 (Direct Cost: ¥6,100,000)
|
|---|
| Keywords | Load Distribvtion / Computer Clvster / Network Computing / Distributed Shared Memory / Distributed Operating System / Cache / Memory Management / Schedvling / 化想化 |
|---|
| Research Abstract |
(1) Research on the Process Fusion and Diffusion for Dynamic Load Balancing.
We have investigated the mechanisms to gather/scatter processing load and to schedule the load under a cluster system consists of several processing nodes, each with different number of processors, different amount of memory and different amount of load but with the same processor architecture. In particular, we have focused on the scheduling algorithm that can realize the Network Supercomputing without making a sacrifice of the real-time response in each node. As such algorithm, we have proposed the Fair Share Priority Scheduler which has both features of the stride scheduling and has low scheduling overhead. On the other hand, we have also developed a runtime system for self-scheduling based dynamic load balance and equipped it with our parallelizing compiler TINPAR.
(2) Research on the distributed operating system for parallel processing.
We have proposed the concept of "Computer Colony" where numbers of independent computers organize a colony on demand and work together as if it is a single SMP system. In order to realize the proposed concept, it is required to provide the computing environment where users, or programs, can have free and easy access, to some extent, to the inexhaustible computing resources on the network without requiring any knowledge about the physical configuration of these computing resources. For this purpose, we have developed a hardware to effectively realize the software based shared memory environment and have proposed a programming interface, "the mission/unit model," which is an extension of the task/thread model in SMP system to meet the various requirements in the network computing environment. And we have made a fundamental design of a distributed operating system "Colonia" which adopts the mission/unit model and implemented it on an software emulator.
|
Report
(3 results)
Research Products
(17 results)
