As business and scientific data have increased dramatically, distributed computing using high-speed networks has become very popular for organizations. From this perspective, this paper is interesting as it presents a security-aware model for such distributed processing, as threats in a distributed environment are a great possibility. This work also addresses handling scheduling issues, using meta-heuristics from other disciplines, for heterogeneous computing and cloud scheduling.
While a homogeneous network of a cluster of computers has seen much research in the area of scheduling of processing tasks, heterogeneous platforms, having different characteristics, have not. This becomes more problematic in a cloud environment as the network is upgraded or extended with a faster processor with better hardware. Since traditional scheduling problems cannot be resolved in realistic time frames, using heuristic algorithms remains the most promising way of providing quality of service. This paper’s main contribution arises from the fact that current hybrid multicore processing nodes consist of both general-purpose central processing units (GPCPUs) and graphic processing units (GPUs). Most of the current research does not address this issue adequately; supercomputing clusters often leave redundant capacity to handle peak demand, thereby rendering them highly underutilized at most other times.
Besides handling the above-mentioned security issues for heterogeneous platforms, the authors propose innovative multidisciplinary approaches to multi-objective optimization (called NSIS) simulating swarm elite intelligence for improving swarm convergence. The work is quite thorough and involved, but could have been more so had the few recent research papers in the field been cited.
The paper makes for an interesting read for those with some experience with optimal distributed processing, especially processors connected by a fast network in a cloud computing environment where having nodes with varying processing powers is expected.
