Scheduling in compute cloud with multiple data banks using divisible load paradigm

The main challenge in a compute cloud system is to design a scheduling strategy for heterogeneous computing resources with shared data banks. The cloud user's job arrives at the Web role, which distributes the load to the worker rules for concurrent processing. The worker role retrieves the respective data from the shared data banks. According to divisible load theory, the scheduling problem is formulated as relevant recursive equations and constraints that are derived from the continuity of processing time due to retrieval from multiple data banks. The scheduling problem in a compute cloud is formulated as a linear programming problem. Finally, we present a satellite image classification problem in a compute cloud as an example to show the adequacy of the proposed solution.