2003 Fiscal Year Final Research Report Summary
Giga-ring cluster with continuously transmitted distributede shared memory system
| Project/Area Number |
14580371
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
計算機科学
|
|---|
| Research Institution | Nagoya Institute of Technology |
|---|
Principal Investigator |
IAWTA Akira Nagoya Institute of Technology, Dept. of Computer Science, Professor, 工学研究科, 教授 (10093098)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
KUROYANAGI Susumu Nagoya Institute of Technology, Dept. of Computer Science, Research Associate, 工学研究科, 助手 (10283475)
MATSUO Hiroshi Nagoya Institute of Technology, Dept. of Computer Science, Professor, 工学研究科, 教授 (00219396)
|
|---|
| Project Period (FY) |
2002 – 2003
|
| Keywords | Network Cluster / Parallel Processing / Distributed Shared Memory |
|---|
| Research Abstract |
Recently, transmission speed between computers becomes more and faster and the network interface which has over 1Gbbp speed has come to be used in general. Therefore, interconnected communication overhead in PC cluster environment consisting of computers connected by an interconnected communication network has been reduced. But communication cost with initial connection stage is still high when accessing the memory of other computers like a conventional distributed shared memory on demand.
This research proposes the new distributed shared memory model that all shared memory contents are always transferred in a cluster system with high transmission speed and no initial communication delay.
We constructed run time library which is an interface to giga-ring shared memory system and giga-ring cluster which has 2, 4, 8 nodes. The specification of each nodes are as follows, CPU : Pentium III x 2 (the aim of one CPU is only transferring data), Memory : 512MB, 1.96, 3.88, 7.68 times (number of CPU' s are 2, 4, 8 respectively) speed up are achieved. Executing FDTD, 1.79, 3.06, 4.03 times (number of CPU' s are 2, 4, 8 respectively) speed up are achieved.
|
Research Products
(2 results)
