It is well known that the main drawback of distributed applications that require high performance is related to the data transfer speed between system nodes. The high speed networks are never enough. The application has to come out with special techniques and algorithms for optimizing data availability. This aspect is increasingly needed for some categories of distributed applications such as computational steering applications, which have to permanently allow users to interactively monitor and control the progress of their applications. Following our previous research which was focused on the development of a set of frameworks and platforms for distributed simulation and computational steering, we introduce in this paper a new model for distributed file systems, supporting data steering, that is able to provide optimization for data acquisition, data output and load balancing while reducing the development efforts, improving scalability and flexibility of the system. Data partitioning is being performed at a logical level allowing multimedia applications to define custom data chunks like frames of a video, phrases of text, regions of an image, etc.
[1]
Olivier Coulaud,et al.
A Steering Environment for Online Parallel Visualization of Legacy Parallel Simulations
,
2006,
2006 Tenth IEEE International Symposium on Distributed Simulation and Real-Time Applications.
[2]
Arun Venkataramani,et al.
Estimating Self-Sustainability in Peer-to-Peer Swarming Systems
,
2010,
Perform. Evaluation.
[3]
Cosmin Marian Poteras,et al.
Grid-Enabled Distributed Simulation - A State Machine Based Approach
,
2010
.
[4]
Yun Tian,et al.
Improving MapReduce performance through data placement in heterogeneous Hadoop clusters
,
2010,
2010 IEEE International Symposium on Parallel & Distributed Processing, Workshops and Phd Forum (IPDPSW).