Load balancing without regret in the bulletin board model

We analyze the performance of protocols for load balancing in distributed systems based on no-regret algorithms from online learning theory. These protocols treat load balancing as a repeated game and apply algorithms whose average performance over time is guaranteed to match or exceed the average performance of the best strategy in hindsight. Our approach captures two major aspects of distributed systems. First, in our setting of atomic load balancing, every single process can have a significant impact on the performance and behavior of the system. Furthermore, although in distributed systems participants can query the current state of the system they cannot reliably predict the effect of their actions on it. We address this issue by considering load balancing games in the bulletin board model, where players can find out the delay on all machines, but do not have information on what their experienced delay would have been if they had selected another machine. We show that under these more realistic assumptions, if all players use the well-known multiplicative weights algorithm, then the quality of the resulting solution is exponentially better than the worst correlated equilibrium, and almost as good as that of the worst Nash. These tighter bounds are derived from analyzing the dynamics of a multi-agent learning system.

[1]  Christos H. Papadimitriou,et al.  Worst-case equilibria , 1999 .

[2]  Vahab S. Mirrokni,et al.  Sink equilibria and convergence , 2005, 46th Annual IEEE Symposium on Foundations of Computer Science (FOCS'05).

[3]  Avrim Blum,et al.  Routing without regret: on convergence to nash equilibria of regret-minimizing algorithms in routing games , 2006, PODC '06.

[4]  G. Piliouras A learning theoretic approach to algorithmic game theory , 2010 .

[5]  Xiaotie Deng,et al.  Settling the Complexity of Two-Player Nash Equilibrium , 2006, 2006 47th Annual IEEE Symposium on Foundations of Computer Science (FOCS'06).

[6]  Y. Mansour,et al.  Algorithmic Game Theory: Learning, Regret Minimization, and Equilibria , 2007 .

[7]  Vahab S. Mirrokni,et al.  Convergence Issues in Competitive Games , 2004, APPROX-RANDOM.

[8]  R. Aumann Subjectivity and Correlation in Randomized Strategies , 1974 .

[9]  Gábor Lugosi,et al.  Prediction, learning, and games , 2006 .

[10]  Sham M. Kakade,et al.  Deterministic calibration and Nash equilibrium , 2004, J. Comput. Syst. Sci..

[11]  Paul G. Spirakis,et al.  The Price of Selfish Routing , 2001, STOC '01.

[12]  D. Fudenberg,et al.  The Theory of Learning in Games , 1998 .

[13]  Paul W. Goldberg,et al.  The Complexity of Computing a Nash Equilibrium , 2009, SIAM J. Comput..

[14]  Paul G. Spirakis,et al.  Atomic Congestion Games: Fast, Myopic and Concurrent , 2008, SAGT.

[15]  Michael Mitzenmacher,et al.  How useful is old information (extended abstract)? , 1997, PODC '97.

[16]  Michael Mitzenmacher,et al.  How Useful Is Old Information? , 2000, IEEE Trans. Parallel Distributed Syst..

[17]  Mohammad Taghi Hajiaghayi,et al.  Regret minimization and the price of total anarchy , 2008, STOC.

[18]  Berthold Vöcking,et al.  Tight bounds for worst-case equilibria , 2002, SODA '02.

[19]  Peter Secretan Learning , 1965, Mental Health.

[20]  Berthold Vöcking,et al.  On the Evolution of Selfish Routing , 2004, ESA.

[21]  Berthold Vöcking,et al.  Fast convergence to Wardrop equilibria by adaptive sampling methods , 2006, STOC '06.

[22]  Éva Tardos,et al.  Multiplicative updates outperform generic no-regret learning in congestion games: extended abstract , 2009, STOC '09.

[23]  Jiantao Wang,et al.  Counter-intuitive throughput behaviors in networks under end-to-end control , 2006, TNET.

[24]  Dean P. Foster,et al.  Regret in the On-Line Decision Problem , 1999 .

[25]  Regret in the On-line Decision Problem 1 , 1999 .

[26]  Paul G. Spirakis,et al.  Atomic Congestion Games: Fast, Myopic and Concurrent , 2009, Theory of Computing Systems.

[27]  P ? ? ? ? ? ? ? % ? ? ? ? , 1991 .

[28]  H PapadimitriouChristos,et al.  The complexity of computing a Nash equilibrium , 2009 .

[29]  H. Peyton Young,et al.  Strategic Learning and Its Limits , 2004 .

[30]  Y. Freund,et al.  Adaptive game playing using multiplicative weights , 1999 .

[31]  Baruch Awerbuch,et al.  Distributed algorithms for multicommodity flow problems via approximate steepest descent framework , 2007, SODA '07.

[32]  R. Vohra,et al.  Calibrated Learning and Correlated Equilibrium , 1996 .