RDIM: a self-adaptive and balanced distribution for replicated data in scalable storage clusters

Authors:
Zhong Liu;Nong Xiao;Xing-Ming Zhou
Affiliations:
Institute of Computer, National University of Defense Technology, Changsha, China;Institute of Computer, National University of Defense Technology, Changsha, China;Institute of Computer, National University of Defense Technology, Changsha, China
Venue:
ISPA'05 Proceedings of the Third international conference on Parallel and Distributed Processing and Applications
Year:
2005

Citing 7
Cited 0

LH*—a scalable, distributed data structure

ACM Transactions on Database Systems (TODS)
Mersenne twister: a 623-dimensionally equidistributed uniform pseudo-random number generator

ACM Transactions on Modeling and Computer Simulation (TOMACS) - Special issue on uniform random number generation
Efficient, distributed data placement strategies for storage area networks (extended abstract)

Proceedings of the twelfth annual ACM symposium on Parallel algorithms and architectures
Compact, adaptive placement schemes for non-uniform requirements

Proceedings of the fourteenth annual ACM symposium on Parallel algorithms and architectures
Design and Implementation of DDH: A Distributed Dynamic Hashing Algorithm

FODO '93 Proceedings of the 4th International Conference on Foundations of Data Organization and Algorithms
Reliability Mechanisms for Very Large Storage Systems

MSS '03 Proceedings of the 20 th IEEE/11 th NASA Goddard Conference on Mass Storage Systems and Technologies (MSS'03)
A Fast Algorithm for Online Placement and Reorganization of Replicated Data

IPDPS '03 Proceedings of the 17th International Symposium on Parallel and Distributed Processing

Quantified Score

Hi-index	0.00

Visualization

Abstract

As storage systems scale from a few storage nodes to hundreds or thousands, data distribution and load balancing become increasingly important. We present a novel decentralized algorithm, RDIM (Replication Under Dynamic Interval Mapping), which maps replicated objects to a scalable collection of storage nodes. RDIM distributes objects to nodes evenly, redistributing as few objects as possible when new nodes are added or existing nodes are removed to preserve this balanced distribution. It supports weighted allocation and guarantees that replicas of a particular object are not placed on the same node. Its time complexity and storage requirements compare favorably with known methods.