A case for redundant arrays of inexpensive disks (RAID)
SIGMOD '88 Proceedings of the 1988 ACM SIGMOD international conference on Management of data
STOC '97 Proceedings of the twenty-ninth annual ACM symposium on Theory of computing
OceanStore: an architecture for global-scale persistent storage
ASPLOS IX Proceedings of the ninth international conference on Architectural support for programming languages and operating systems
DPFS: A Distributed Parallel File System
ICPP '02 Proceedings of the 2001 International Conference on Parallel Processing
PAST: A Large-Scale, Persistent Peer-to-Peer Storage Utility
HOTOS '01 Proceedings of the Eighth Workshop on Hot Topics in Operating Systems
SOSP '03 Proceedings of the nineteenth ACM symposium on Operating systems principles
Weighted distributed hash tables
Proceedings of the seventeenth annual ACM symposium on Parallelism in algorithms and architectures
SAN Optimal Multi Parameter Access Scheme
ICNICONSMCL '06 Proceedings of the International Conference on Networking, International Conference on Systems and International Conference on Mobile Communications and Learning Technologies
Dynamo: amazon's highly available key-value store
Proceedings of twenty-first ACM SIGOPS symposium on Operating systems principles
Hi-index | 0.00 |
We consider an optimisation problem which is motivated from storage virtualisation in the Internet. While storage networks make use of dedicated hardware to provide homogeneous bandwidth between servers and clients, in the Internet, connections between storage servers and clients are heterogeneous and often asymmetric with respect to upload and download. Thus, for a large file, the question arises how it should be fragmented and distributed among the servers to grant "optimal" access to the contents. We concentrate on the transfer time of a file, which is the time needed for one upload and a sequence of n downloads, using a set of m servers with heterogeneous bandwidths. We assume that fragments of the file can be transferred in parallel to and from multiple servers. This model yields a distribution problem that examines the question of how these fragments should be distributed onto those servers in order to minimise the transfer time. We present an algorithm, called FLOWSCALING, that finds an optimal solution within running time O(mlogm). We formulate the distribution problem as a maximum flow problem, which involves a function that states whether a solution with a given transfer time bound exists. This function is then used with a scaling argument to determine an optimal solution within the claimed time complexity.