Cluster technology that enables a group of computers working closely to form a single computer, has been a booming research field in computer engineering and network. State of the art cluster applications, such as large scale database processing, require thousands of computers to work together. However, current solutions for cluster database applications require some kinds of centralized control, such as center indexing or localization mapping. When the team sizes of those clusters scale up, centralized control will inevitably bring on the bottle-neck in network and processor traffic. In this paper, we put forward a decentralized algorithm for data processing in database cluster. The key is an intelligent routing algorithm to route data to the exact server without knowing any global knowledge about the cluster. It is able to efficiently route requests through the cluster via the use of local decision theoretic models. Moreover, each processing request is used to improve the routing of other request leading to a dramatic performance improvement.
[1]
Sang-Hwa Chung,et al.
Information Retrieval on an SCI-Based PC Cluster
,
2001,
The Journal of Supercomputing.
[2]
James Annis.
Large scale cluster surveys and distributed computing
,
2002
.
[3]
Duncan J. Watts,et al.
Collective dynamics of ‘small-world’ networks
,
1998,
Nature.
[4]
Rajkumar Buyya,et al.
High Performance Cluster Computing: Architectures and Systems
,
1999
.
[5]
Marie desJardins,et al.
Agent-organized networks for dynamic team formation
,
2005,
AAMAS '05.
[6]
Yang Xu,et al.
An integrated token-based algorithm for scalable coordination
,
2005,
AAMAS '05.
[7]
Mark Baker,et al.
Cluster Computing White Paper
,
2000,
ArXiv.