A Hierarchical Load Balancing Strategy Considering Communication Delay Overhead for Large Distributed Computing Systems

Table 1

Load balancing overhead and memory overhead on 32 K emulated processors.

GreedyCommLB	Number of objects
	128 k	256 k	512 k	1 m
Centralized
Balancing overhead	10.2961 s	21.1509 s	38.1038 s	—
Memory overhead	41.4052 M	82.8662 M	165.8105 M	—
Hierarchical
Balancing overhead at level 0	0.1958 s	0.4399 s	0.7824 s	1.7164 s
Memory overhead at level 0	5.1542 M	10.3281 M	24.9335 M	41.4052 M
Balancing overhead at level 1	0.4616 s	1.2910 s	3.6935 s	5.2248 s
Memory overhead at level 1	10.3281 M	24.9335 M	41.4052 M	82.8662 M