Table 1.
Network size and computational environment of previous studies in closed queueing networks.
Fig 1.
Flowchart of the parallel computation algorithm for Mean Value Analysis (MVA) using Message Passing Interface (MPI).
Fig 2.
Flowchart of the event-driven parallel simulation algorithm for BCMP queueing networks.
Table 2.
Massively parallel computing environment for MVA.
Table 3.
Results with 128 parallels and various values of N, C, and K.
Fig 3.
Change in computation time for an increase in number of nodes N and number of people in system K (C = 3).
Fig 4.
Variation of computation time with respect to N (K = 500,C = 3).
Fig 5.
Variation of computation time with respect to K (N = 33,C = 3).
Table 4.
Computation time with different number of parallels (N = 33,C =3,K = 500).
Table 5.
Simulation accuracy for varying C when N = 33,K = 500.
Table 6.
Simulation accuracy for varying N and K when C = 3 (Eq(12)).
Table 7.
The change in the value of Eq (10) for the number of classes C ≥ 4 (N = 33, K = 500, ε = 0.01).
Fig 6.
Root-mean-square error of simulation relative to theoretical value, plotted against the mean of the standard deviation of the number of customers in the mean system for each simulation.