Model for Simulation of Heterogeneous High-Performance Computing Environments

Abstract

Abstract. This paper proposes a new model to predict the process execution behavior on heterogeneous multicomputing environments. This model considers the process execution costs such as processing, hard disk acessing, message transmitting and memory allocation. A simulator of this model was developed which help to predict the execution behavior of processes on distributed environments under different scheduling techniques. Besides the simulator, it was developed a suite of benchmark tools in order to parameterize the proposed model with data collected from real environments. Experiments were conduced to evaluate the proposed model which used a parallel application executing on a heterogeneous system. The obtained results show the model ability to predict the actual system performance, providing an useful model for developing and evaluating techniques for scheduling and resource allocation over heterogeneous and distributed systems.

Grid job scheduling using Route with Genetic Algorithm support

“…In [13], Mello and Senger propose a mathematical model to predict the execution behavior of parallel applications on heterogeneous computing environments. Such model was implemented in a simulator (SchedSim.…”

“…5 The grid environment was organized in local networks composed of up to five computers. Within the same network, computers present a latency (RTT-Round-Trip Time, according to the fixed model by Hockney [22]) of 0.0001 (mean value extracted by the network benchmark by Mello and Senger [13] for a Gigabit Ethernet infrastructure). In order to characterize the latency among computers of different networks, experiments have been carried out on local, metropolitan and worldwide networks.…”

“…C j is defined in (11) where PM j is the total processing cost of the task at index j , CAP d is the capacity of the computer d where the task j is allocated, and CC j is the communication cost of the task at index j to any other task (from the same parallel application) running on other computers. CC j is defined in (13) where CPS(i j , i k ) is the communication cost between the tasks at indexes j and k of the chromosome i which run, respectively, on the computers i j and i k .…”

“…These measures were extracted by using the benchmark proposed by Mello and Senger [13]. 5 The grid environment was organized in local networks composed of up to five computers.…”

“…The processing costs are converted to a unique measure for all environment computers. For this conversion, we used the benchmark proposed by Mello and Senger [13] which allows to obtain the CPU capacity in MIPS (million of instruction per second).…”