Summary New generation application problems in big data and high‐performance computing (HPC) areas claim very diverse operational properties. The convergence requires the dynamic behavior of system components. Load balancing is a critical issue in response to the highly unpredictable, dynamic, and data‐oriented behavior of the system. Possible practical constraints such as communication and load transfer delays play an essential role in designing a dynamic load balancer. On the other hand, according to most of the new platforms' distributed nature, the load balancer should be able to perform in a fully distributed manner. In this research, we consider practical issues, including different processing power, storage capability, communication, load transfer delays, and propose two distributed and optimized load balancing methods in HPC for Big Data processing. We model the constraints and present an argument named compensating factor for the optimized load balancer. We try to minimize the task execution time by reducing the nodes' idle time. We evaluate the proposed methods in different scenarios by using Monte Carlo. Evaluations results show that proposed methods decrease idle time significantly while being scalable to network size and applicable in heterogeneous networks with dynamic resources and configuration.

SUMMARY The complexity of computing systems introduces a few issues and challenges such as poor performance and high energy consumption. In this paper, we first define and model resource contention metric for high performance computing workloads as a performance metric in scheduling algorithms and systems at the highest level of resource management stack to address the main issues in computing systems. Second, we propose a novel autonomic resource contention‐aware scheduling approach architected on various layers of the resource management stack. We establish the relationship between distributed resource management layers in order to optimize resource contention metric. The simulation results confirm the novelty of our approach.Copyright © 2013 John Wiley & Sons, Ltd.

A new computational technique for training of multilayer feedforward neural networks with sigmoid activation function of the units is proposed. The proposed algorithm consists two phases. The first phase is an adaptive training step calculation, which implements the steepest descent method in the weight space. The second phase is estimation of calculated training step rate, which reaches a state of activity of the units for each training iteration. The simulation results are provided for the test example to demonstrate the efficiency of the proposed method, which solves the problem of training step choice in multilayer perceptrons.