Exploiting biased load information in direct-neighbour load balancing policies

Massively parallel architectures require distributed load balancing policies based on local load information only. This may sometime limit their effectiveness. The paper presents a class of direct-neighbour load balancing policies that aim to overcome the limits of locality by introducing biases in the load information exchanged between the system nodes. This permits load information to keep into account a more global view of the system. The efficiency of the presented policies is evaluated in dependence of the characteristics of the system load. Experimental results show that the transmission of biased load information provides high efficiency unless the dynamicity of the load becomes too high, in which case it is preferable to exploit non-biased load information. (C) 1999 Elsevier Science B.V. All rights reserved.