Scalable load balancing mechanism by asynchronous migrations

Abstract

We propose a scalable load balancing mechanism based on asynchronous migration. For a load balancing mechanism to be scalable, the communication overhead must be insensitive to the size of the systems. The previous models migrate processes synchronously between senders and receivers. We identified the massive communication overhead in the previous model due to the synchrony. In the proposed model, the senders release the saturated processes on the sharable bulletin boards, and the receivers draw the released processes depending on the local load condition. The asynchronous migration takes advantages in reducing the communication overhead and supporting the interoperability among the heterogeneous management principles. The goal of the paper is to investigate the behavior of the asynchronous migration in the sacalable load balancing mechanism. Keywords: Load Balancing, Asynchronous Migration, Bulletin Boards, Client-Solicitor Interactions