1: \begin{abstract}

2:   In this paper we consider neighborhood load balancing in the

3:   context of selfish clients.  We assume that a network of $n$

4:   processors and $m$ tasks is given.  The processors may have

5:   different speeds and the tasks may have different weights.  Every

6:   task is controlled by a selfish user. The objective of the user is

7:   to allocate his/her task to a processor with minimum load.

8: 

9:   We revisit the concurrent probabilistic protocol introduced in

10:   \cite{Berenbrink2011}, which works in sequential rounds.  In each

11:   round every task is allowed to query the load of one randomly chosen

12:   neighboring processor. If that load is smaller the task will

13:   migrate to that processor with a suitably chosen probability.  Using

14:   techniques from spectral graph theory we obtain upper bounds on the

15:   expected convergence time towards approximate and exact Nash

16:   equilibria that are significantly better than the previous results

17:   in \cite{Berenbrink2011}.  We show results for uniform tasks on

18:   non-uniform processors and the general case where the tasks have

19:   different weights and the machines have speeds.  To the best of our

20:   knowledge, these are the first results for this general setting.

21: \end{abstract}

22: