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

Table 2

Load balancing overhead and memory overhead on 64 K emulated processors.

GreedyCommLB	Number of objects
	128 k	256 k	512 k	1 m
Centralized
Balancing overhead	10.84 s	25.22 s	45.18 s	—
Memory overhead	41.40 M	82.87 M	165.81 M	—
Hierarchical
Balancing overhead at level 0	0.42 s	1.08 s	3.95 s	4.95 s
Memory overhead at level 0	10.33 M	24.93 M	41.40 M	82.87 M
Balancing overhead at level 1	0.49 s	1.03 s	2.30 s	5.63 s
Memory overhead at level 1	12.45 M	20.68 M	41.41 M	82.87 M