Software defined networking (SDN) is currently regarded as one of the most promising paradigms of future Internet. Although the availability and scalability that a single and centralized controller suffers from could be alleviated by using multiple controllers, there lacks a flexible mechanism to balance load among controllers. This paper proposes a load balancing mechanism based on a load informing strategy for multiple distributed controllers. With the mechanism, a controller can make load balancing decision locally as rapidly as possible. Experiments based on floodlight show that our mechanism can balance the load of each controller dynamically and reduce the time of load balancing.
The transmission-line modeling (TLM) used for nonlinear finite-element (FE) solution has a paramount feature that the admittance matrix is unchanged and only needs one-time factorization; and this feature becomes a drawback when the required number of TLM iterations increase due to the mismatch between the transmission-line impedance and the load. In this paper, a matrix-free TLM scheme is proposed to make use of the solved nonlinear reluctivities without employing any matrices at each timestep, thus substantially decreasing the number of required TLM iterations. The matrix-free solver is suitable for massively parallel processing and the design is implemented on the Tesla V100 graphics processing unit (GPU). A speedup of more than 27 times is obtained compared with a commercial FE package for different problem sizes while maintaining high accuracy.
In this article, a novel edge-domain decomposition (EDD) method is proposed to solve 3-D nonlinear finite element (FE) problems of electromagnetic devices and transient field circuit co-simulation. The method applies reduced magnetic vector potential formulation to discretize the physical problem based on 3-D edge elements, and the solution region is divided into many sub-domains that only contain one edge unknown. The solution of lightweight nonlinear sub-domain systems can be massively parallelized, and the neighbor-to-neighbor communication scheme eliminates the need to assemble the global FE matrix. This article also introduces an indirect coupling scheme to handle large eddy currents to interface the EDD FE system with external circuits. The abovementioned algorithms are then implemented on a many-core GPU for transient field circuit co-simulation. The result shows an auto-gauging property, and the comparison with a commercial FE software indicates a speedup of over 43 times with relative error less than 2%.Index Terms-3-D edge element, domain decomposition, Eddy current field, field-circuit coupling, finite element (FE) method, graphics processors, nonlinear, parallel processing, reduced magnetic vector potential.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.