Two-Server Queuing Model for a Community College

I investigate whether a G/G/2 queuing model can be used under the situation that both interarrival and service times are not exponentially distributed. Secondly I examine whether any superior routing policy exists in the case that two servers have different service rates. For the first question, I found that the empirically measured interarrival and service times were not exponentially distributed but stable when I used the CUSUM method. For the second question, I used ANOVA method and mean difference tests. I found that some differences among routing policies appeared when two servers had different service rates. Based on the results, I can recommend manager to use most-productive-first routing policy provided that two servers has different service rates.