We extend the lower bound of Adler et. al [1] and Berenbrink [2] for parallel randomized load balancing algorithms.
The setting in these asynchronous and distributed algorithms is of n balls and n bins. The algorithms begin by each ball choosing d bins independently and uniformly at random. The balls and bins communicate to determine the assignment of each ball to a bin. The goal is to minimize the maximum load, i.e., the number of balls that are assigned to the same bin. In [1,2], a lower bound of $\Omega(\sqrt[r]{ \log n / \log \log n})$ is proved if the communication is limited to r rounds.
Three assumptions appear in the proofs in [1,2]: the topological assumption, random choices of confused balls, and symmetry. We extend the proof of the lower bound so that it holds without these three assumptions. This lower bound applies to every parallel randomized load balancing algorithm we are aware of [1,2,3,4].
[1]
Volker Stemann,et al.
Parallel balanced allocations
,
1996,
SPAA '96.
[2]
Moti Medina,et al.
Revisiting Randomized Parallel Load Balancing Algorithms
,
2009,
SIROCCO.
[3]
Eli Upfal,et al.
Balanced Allocations
,
1999,
SIAM J. Comput..
[4]
Friedhelm Meyer auf der Heide,et al.
Allocating Weighted Jobs in Parallel
,
1999,
Theory of Computing Systems.
[5]
Berthold Vöcking,et al.
How asymmetry helps load balancing
,
1999,
JACM.
[6]
Martin Raab,et al.
"Balls into Bins" - A Simple and Tight Analysis
,
1998,
RANDOM.
[7]
M. Mitzenmacher,et al.
Parallel randomized load balancing
,
1998
.
[8]
Friedhelm Meyer auf der Heide,et al.
Contention Resolution in Hashing Based Shared Memory Simulations
,
2000,
SIAM J. Comput..