The Annals of Applied Probability

Two-server closed networks in heavy traffic: diffusion limits and asymptotic optimality

Sunil Kumar

Full-text: Open access


One of the successes of the Brownian approximation approach to dynamic control of queueing networks is the design of a control policy for closed networks with two servers by Harrison and Wein. Adopting a Brownian approximation with only heuristic justification, theyinterpret the optimal control policy for the Brownian model as a static priority rule and conjecture that this priority rule is asymptotically optimal as the closed networks’s population becomes large. This paper studies closed queueing networks with two servers that are balanced, that is, networks that have the same relative load factor at each server. The validity of the Brownian approximation used by Harrison and Wein is established by showing that, under the policy they propose, the diffusion-scaled workload imbalance process converges weakly in the infinite population limit to the diffusion predicted by the Brownian approximation. This is accomplished by proving that the fluid limits of the queue length processes undergo state space collapse in finite time under the proposed policy, thereby enabling the application of a powerful new technique developed by Williams and Bramson that allows one to establish convergence of processes under diffusion scaling by studying the behavior of limits under fluid scaling. A natural notion of asymptotic optimality for closed queueing networks is defined in this paper.The proposed policy is shown to satisfy this definition of asymptotic optimality by showing that the performance under the proposed policy approximates bounds on the performance under every other policy arbitrarily well as the population increases without bound.

Article information

Ann. Appl. Probab. Volume 10, Number 3 (2000), 930-961.

First available in Project Euclid: 22 April 2002

Permanent link to this document

Digital Object Identifier

Mathematical Reviews number (MathSciNet)

Zentralblatt MATH identifier

Primary: 60K25: Queueing theory [See also 68M20, 90B22]
Secondary: 60F17: Functional limit theorems; invariance principles 90F35

Queueing networks functional limit theorems diffusion limits fluid limits scheduling policies


Kumar, Sunil. Two-server closed networks in heavy traffic: diffusion limits and asymptotic optimality. Ann. Appl. Probab. 10 (2000), no. 3, 930--961. doi:10.1214/aoap/1019487514.

Export citation


  • [1] Berman, A. and Plemmons, R. (1994). Nonnegative Matrices in the Mathematical Sciences.
  • SIAM, Philadelphia.
  • [2] Billingsley, P. (1968). Convergence of Probability Measures. Wiley, New York.
  • [3] Bramson, M. (1998). State space collapse with application to heavytraffic limits for multiclass queueing networks. Queueing Sys. 30 89-148.
  • [4] Chen, H. and Mandelbaum, A. (1990). Leontief Systems, RBVs and RBMs. In Applied Stochastic Analysis (M. H. A. Davis and R. J. Elliott, ed.) 1-43. Gordon and Breach, Langhorne, PA.
  • [5] Dai, J. and Dai, W. (1998). A heavytraffic limit theorem for a class of open queueing networks with finite buffers. Preprint.
  • [6] Dai, J. and Williams, R. (1995). Existence and uniqueness of semimartingale reflecting Brownian motions in convex polyhedrons. Theory Probab. Appl. 40 3-53.
  • [7] Dai, J. G. (1995). On Positive Harris recurrence of multiclass queueing networks: a unified approach via fluid limit models. Ann. Appl. Probab. 5 49-77.
  • [8] Dai, J. G. and Weiss, G. (1996). Stabilityand instabilityof fluid models for reentrant lines. Math. Oper. Res. 21 115-134.
  • [9] Ethier, S. N. and Kurtz, T. G. (1986) Markov Processes: Characterization and Convergence. Wiley, New York.
  • [10] Harrison, J. M. (1985). Brownian Motion and Stochastic Flow Systems. Wiley, New York.
  • [11] Harrison, J. M. (1988). Brownian models of queueing networks with heterogeneous customer populations. In Stochastic Differential Systems, Stochastic Control Theory and Applications (W. Fleming and P. L. Lions, eds.) 147-186. Springer, New York.
  • [12] Harrison, J. M. and Van Mieghem, J. A. (1997). Dynamic control of Brownian networks: state space collapse and equivalent workload formulations. Ann. Appl. Probab. 7 747-771.
  • [13] Harrison, J. M. and Wein, L. M. (1989). Scheduling network of queues: heavytraffic analysis of a simple open network. Queueing Sys. 5 265-280.
  • [14] Harrison, J. M. and Wein, L. M. (1990). Scheduling networks of queues: heavytraffic analysis of a two-station closed network. Oper. Res. 38 1052-1064.
  • [15] Iglehart, D. L. and Whitt, W. (1970). Mutliple channel queues in heavytraffic I. Ann. Appl. Probab. 2 150-177.
  • [16] Jin, H., Ou, J., and Kumar, P. R. (1997). The throughput of closed queueing networks: functional bounds, asymptotic loss, efficiency and the Harrison-Wein conjectures. Math. Oper. Res. 22 886-920.
  • [17] Kumar, S. and Kumar, P. R. (1994). Performance bounds for queueing networks and scheduling policies. IEEE Trans. Automat. Control. 39 1600-1611.
  • [18] Kumar, S. and Kumar, P. R. (1996). Fluid limits and the efficiencyof scheduling policies for stochastic closed reentrant lines in heavytraffic. Stochastic Networks: Stability and Rare Events. Lecture Notes in Statist. 117 41-64. Springer, Berlin.
  • [19] Lindvall, T. (1992). Lectures on the Coupling Method. Wiley, New York.
  • [20] Martins, L. F., Shreve, S. E. and Soner, H. M. (1996). Heavytraffic convergence of a controlled multiclass queueing network. SIAM. J. Control Optim. 34 2133-2171.
  • [21] Solberg, J. J. (1981). Capacityplanning using a stochastic workflow model. AIIE Trans. 13 116-122.
  • [22] Spearman, M. L., Woodruff, D. L. and Hopp, W. J. (1990). CONWIP: a pull alternative to Kanban. Intl. J. Production Res. 28 879-894.
  • [23] Williams, R. (1998). Diffusion approximations for open multiclass queueing networks: sufficient conditions involving state space collapse. Queueing Sys. 30 27-88.
  • [24] Williams, R. (1998). An invariance principle for semimartingale reflecting Brownian motions in an orthant. Queueing Sys. 30 5-25.