A Note on the Accuracy of Several Existing Approximations for M/Ph/m Queues

High variability of system parameters is a complicating factor in the modeling of the performance of big data systems. In this paper, we assess the potential inaccuracy of several existing approximations for evaluating the mean number of jobs queued in a parallelized device that can be represented as an M/Ph/m queue. Unlike existing studies, we consider the effect of the third moment of the service time, or equivalently, its skewness. We show that the approximations accuracy can be poor even for "easy" examples with a low coefficient of variation of the service time. Our examples demonstrate the important influence of the skewness of the service time distribution on the accuracy of the approximations. None of the approximations accounts for this property. We provide recommendations for the choice of the approximation that allow the user to choose the best suited approximation based on the actual queue parameters.

[1]  George P. Cosmetatos.  Some Approximate Equilibrium Results for the Multi-Server Queue (M/G/r)* , 1976 .

[2]  Toshikazu Kimura,et al.  Approximations for multi-server queues: System interpolations , 1994, Queueing Syst. Theory Appl..

[3]  Toshikazu Kimura.  A two-moment approximation for the mean waiting time in the GI/G/s queue , 1985 .

[4]  Henk Tijms,et al.  Stochastic modelling and analysis: a computational approach , 1986 .

[5]  Mor Harchol-Balter,et al.  Closed form solutions for mapping general distributions to quasi-minimal PH distributions , 2006, Perform. Evaluation.

[6]  A. Horváth,et al.  Matching Three Moments with Minimal Acyclic Phase Type Distributions , 2005 .

[7]  James Martin,et al.  Systems analysis for data transmission , 1972 .

[8]  A. M. Lee,et al.  Queueing Processes Associated with Airline Passenger Check-in , 1959 .

[9]  Toshikazu Kimura.  A two-moment approximation for the mean waiting time in the GI/G/s queue , 1985 .

[10]  Gunter Bolch,et al.  Queueing Networks and Markov Chains , 2005 .

[11]  O. J. Boxma,et al.  Approximations of the Mean Waiting Time in an M/G/s Queueing System , 1979, Oper. Res..

[12]  Mor Harchol-Balter,et al.  The effect of higher moments of job size distribution on the performance of an M/G/s queueing system , 2007, PERV.

[13]  Ronald W Wolff,et al.  The Effect of Service Time Regularity on System Performance. , 1977 .

[14]  Arnold O. Allen,et al.  Probability, statistics and queueing theory - with computer science applications (2. ed.) , 1981, Int. CMG Conference.

[15]  Ward Whitt,et al.  The effect of variability in the GI/G/s queue , 1980, Journal of Applied Probability.

[16]  Thomas Begin,et al.  Considerations in Workload Characterization for Parallel Access Volumes , 2009, Int. CMG Conference.