Estimating Effective Capacity in Erlang Loss Systems under Competition

Authors:
Andrew M. Ross;J. George Shanthikumar
Affiliations:
Department of Industrial and Systems Engineering, Lehigh University, Bethlehem, USA 18015;Department of Industrial Engineering and Operations Research, University of California Berkeley, Berkeley, USA 94720
Venue:
Queueing Systems: Theory and Applications
Year:
2005

Citing 9
Cited 1

The queue inference engine: deducing queue statistics from transactional data

Management Science
The pointwise stationary approximation for M1/M1/s

Management Science
The pointwise stationary approximation for queues with nonstationary arrivals

Management Science
Some effects of nonstationarity on multiserver Markovian queueing systems

Operations Research
Regular Article: Derivatives of the Matrix Exponential and Their Computation

Advances in Applied Mathematics
Sensitivity to the service-time distribution in the nonstationary Erlang loss model

Management Science
Expokit: a software package for computing matrix exponentials

ACM Transactions on Mathematical Software (TOMS)
Improving the Sipp Approach for Staffing Service Systems That Have Cyclic Demands

Operations Research
Introduction to Probability Models, Ninth Edition

Introduction to Probability Models, Ninth Edition

OM Practice---Work Expands to Fill the Time Available: Capacity Estimation and Staffing Under Parkinson's Law

Manufacturing & Service Operations Management

Quantified Score

Hi-index	0.00

Visualization

Abstract

We consider an Erlang loss system (modem bank) with two streams of arriving customers, where arrival rates vary by time-of-day. We can observe one of the traffic streams (our customers), but we do not know how many servers the system has, or the characteristics of the other stream. Using detailed sample-path data, we construct a maximum likelihood estimator that makes good use of the data, but is slow to evaluate. As an alternative, we present an estimation system based on traffic data summarized by hour. This estimation system is much faster, and tends to produce good lower bounds on the size of the system and competing traffic.