BLuRGER: balancing load using request graph early reservation (システムソフトウェアとオペレーティング・システム)
スポンサーリンク
概要
- 論文の詳細を見る
This paper examines optimizations of file placement and load balancing on an asymmetric cluster-based file system by means of early reservations of a set of files to be requested by the client. A client defines a request graph for a set of files based on file attributes and relations, the structure of the graph is relevant to expected request patterns, which in turn have an effect on the load distribution over the cluster nodes. The server receives from the client reservations in advance to make early placement decisions in order to maintain a balanced cluster workload while satisfying the request graph structural properties and resource requirements. In this work, a model for the client request graph and a corresponding server decision algorithm are defined and discussed. The effect of this approach is simulated using variable parameters and compared against classical load balancing approaches.
- 2009-01-21
著者
-
Honsali Khalil
Nagoya Institute Of Technology
-
Tsumura Tomoaki
Nagoya Institute Of Technology
-
SAITO SHOICHI
Nagoya Institute of Technology
-
MATSUO HIROSHI
Nagoya Institute of Technology
関連論文
- An Efficient Analysis of Worst Case Flush Timings for Branch Predictors(Processor Architecture)
- BLuRGER: balancing load using request graph early reservation (システムソフトウェアとオペレーティング・システム)
- An Efficient Analysis of Worst Case Flush Timings for Branch Predictors
- An Efficient Analysis of Worst Case Flush Timings for Branch Predictors