Genetic algorithm optimized receiver-oriented packet clustering in multi-buffer network card

In a network card, there is a packet clustering mechanism that if optimized, will increase the speed of packet reading/access at the application layer. This packet clustering optimization is an expansion of our previous base network card optimization in cloud network environment, using genetic algorithm. The expansion adds a new gene to our chromosome, which is number of buffers, and its effect towards convergence of fitness values is also a part of our objectives. Its optimization factors revolve around the number of buffers, the number of applications that need access to the clusters, capacity of each buffer, transmission size, packet size, and number of packets. All of these properties determine the duration of packet reading of the application layer; the shorter the better. This packet reading/access duration will be added to the duration of network card�s packet processing, thus the orientation of throughput is towards the application layer. Our simulations work on multiple network cards simultaneously and the resulted optimal solutions prove to accelerate both the network card packet processing and its subsequent packet reading by the application layer, as a result of optimal packet clustering. © 2016 American Scientific Publishers. All rights reserved.