pratikkejriwal
commented
6 years ago problem statement defined as: any computation that has to be done on the cloud, will take a long time even with higher computation speed of the cloud and that is because of the propagation time to and from the cloud and the client. The nodes are at a variable distance from the client. Traditionally the task would have been sent to the readily available fog node. But as discussed that would not be the best fit, alternately the node at the shortest distance from the client can be chosen, but it also creates the processing time problem. As the total time for computation is evaluated by using the equation: totalt = (2 * propagationt)+ processingt (1) where totalt is the total time taken for the computation while propagationt is the time taken for the data to travel from client to the node (considering that the travel time to and from the client to node as well as node to client is same) and processingt is the time taken by the processor to compute the task. Though the propagation time which accounts for a large computation time is minimized, the processing time is not reduced which is of major concern. Now by applying Scatter-Gather algorithm [1-6], the decomposition of data is performed and the decomposed data is distributed such that the processing time (processingt) is reduced. This algorithm though looks good on paper may incur some latency as the tasks are divided based on their priority and not based on their individual distances with the nodes, which in turn might result in the propagation delay thereby increasing propagationt which will result in increased totalt. So it becomes clear that in order to reduce the totalt we not only need to optimize propagationt or processingt but we have to optimize both of them.
The issue is quite obvious. Elaboration required. Try to reuse the existing template