Anitha senthil
Newbie level 1
Basically I focus on grouping “nearby” grid nodes into clusters and computing the “distance” between these clusters. This information is sufficient for a scheduler to determine which groups of nodes can be efficiently employed to solve smaller problems, and which combinations of clusters are likely to be effective for larger problems.
Initially I form the clusters of compute nodes based on RTT between the node and the landmark(set of reference nodes).
If that clustering of nodes will be stable or it can be dynamic(based on RTT) ?
If the clustering of grid nodes will be dynamic,how the grid scheduler handles the scheduling of jobs to dynamic clusters?
consider the following scenario:
Initially, clusters of grid nodes are formed. scheduler allocated jobs to those clusters. suddenly due to the dynamic nature, clusters are again reconstructed. At that time, what about the status of the jobs allocated to the previous clusters and how it is handled?
Another query is if the clusters are dynamic, when the reconstruction of clusters is to be done? Is there any constraint or periodically updated?
Initially I form the clusters of compute nodes based on RTT between the node and the landmark(set of reference nodes).
If that clustering of nodes will be stable or it can be dynamic(based on RTT) ?
If the clustering of grid nodes will be dynamic,how the grid scheduler handles the scheduling of jobs to dynamic clusters?
consider the following scenario:
Initially, clusters of grid nodes are formed. scheduler allocated jobs to those clusters. suddenly due to the dynamic nature, clusters are again reconstructed. At that time, what about the status of the jobs allocated to the previous clusters and how it is handled?
Another query is if the clusters are dynamic, when the reconstruction of clusters is to be done? Is there any constraint or periodically updated?
