| Abstract |
The continuous growth of electronic payment methods in business transactions is a reality that boosts e-commerce, being present more and more in our daily lives. Considering this situation, we proposed in a previous work a model denoted GetLB, which contains a scheduler that provides good results when compared to the traditional dispatching approach – the Round-Robin. Although offering a good distribution of transactions to processing (PMs), GetLB’s scheduling routine is time consuming since each input is always analyzed against each target PM. Thus, in this paper we are proposing GetLB++ – a GetLB im- provement that covers scheduling computation efficiency by bursting a transaction to a specific PM in accordance with a updated in-memory decreasing-sorted list of PM capacities. The results using an Amazon EC2 cluster instance showed a higher scheduling speed on GetLB++ in comparison with the standard GetLB, presenting gains about 20% on the makespan time. Besides scalability on EFT systems, GetLB++’s contributions are not limited to the context of transactional systems, but can also be extended for load balancing in e-commerce systems, cloud computing, and parallel programming. |
