Abstract
We consider estimation problems in G/G/∞ queue under incomplete information. Specifically, we are interested in scenarios where it is infeasible to track each individual job in the system and only aggregate statistics are known or observable. We first show that the minimum expected square estimator for the queue length process can be written as the sum of the minimum square estimators for an indicator function associated with each job, which is simply the survival function of the service time variable for each job. We also obtain tight lower and upper bounds on the time average of the square estimation error. Next we look at the inverse problem of estimating the service time distribution when the observed process is only the queue length process. We develop an on-line stochastic-optimization based estimation algorithm for the service time distribution and study its convergence under different parameter settings.
- N. H. Bingham and Susan M. Pitts. Non-Parametric Estimation for the M/G/∞ Queue. Ann. Inst. Statis. Math., 51(1):71--97, 1999.Google ScholarCross Ref
- M. Brown. An M/G/∞ Estimation Problem. Ann. Math. Statist, 41:651--654, 1970.Google ScholarCross Ref
- Hanhua Feng and Vishal Misra. On the Relationship Between Coefficient of Variation and Performance of M/G/1 -- FB Queues. SIGMETRICS Performance Evaluation Review, 32(2):17--19, 2004. Google ScholarDigital Library
- Rudolf Grubel and Susan M. Pitts. Non-Parameteric Estimation in Renewal Theory, I: the empirical renewal function. The Annals of Statistics, 21:1431--1451, 1993.Google ScholarCross Ref
- Richard C. Larson. The queue inference engine: Deducing queue statistics from transactional data. Management Science, 36(5):586--601, 1990. Google ScholarDigital Library
- A. Nozari and Whitt W. Estimating Average Production Intervals Using Inventory Measurements: Little's Law for Partially Observable Processes. Operations Research, 36:308--323, 1988. Google ScholarDigital Library
- J. Pickands and R. A. Stine. Estimation for an M/G/∞ Queue with Incomplete Information. Biometrika, 84:295--308, 1997.Google ScholarCross Ref
Recommendations
The M/G/1+G queue revisited
We consider an M/G/1 queue with the following form of customer impatience: an arriving customer balks or reneges when its virtual waiting time, i.e., the amount of work seen upon arrival, is larger than a certain random patience time. We consider the ...
Tail asymptotics for the queue length in an M/G/1 retrial queue
In this paper, we study the tail behavior of the stationary queue length of an M/G/1 retrial queue. We show that the subexponential tail of the stationary queue length of an M/G/1 retrial queue is determined by that of the corresponding M/G/1 queue, and ...
Bounds for Some Generalizations of the GI/G/1 Queue
<P>Expressions are derived for the expected number and expected wait in queue for the following generalizations of the GI/G/1 queue: 1 arrivals in batches of random size, 2 service in batches of fixed size, 3 queues with added delay for the first ...
Comments