Table 2: Experimental results for linear ELP M = 5 facilities l max(M) - arrival rate for the busiest facility EFC Gap Coverage Radius Beta(2,2) Beta(.5,.5) Beta(.25,2) Uniform 0.110 0.253 0.271 0.718 0.2 0.263 0.353 2.590 0.120 0.245 0.244 0.654 0.227 0.221 2.271 0.130 0.240 0.225 0.571 0.200 0.126 1.853 0.150 0.228 0.205 0.365 0.140 0.024 0.824 0.175 0.217 0.311 0.084 0.000 0.554 0.277 0.023 0.385 0.300 0.232 0.159 0.400 0.226 0.132 0.500 0.220 0.099 0.600 0.211 0.055 M = 10 facilities Arrival rate for the busiest facility 0.158 0.599 0.1 0.324 0.575 4.987 0.060 0.125 0.128 0.527 0.254 0.282 4.273 0.065 0.122 0.109 0.406 0.219 0.086 3.062 0.075 0.115 0.101 0.146 0.010 1.632 0.088 0.112 0.100 0.207 0.121 1.073 0.178 0.098 0.779 0.102 0.136 0.019 0.360 0.224 0.250 0.147 M = 20 facilities 0.028 0.077 0.081 0.05 0.535 0.627 3.645 0.030 0.071 0.062 0.144 0.418 0.241 1.878 0.033 0.057 0.145 1.245 0.038 0.050 0.008 0.768 0.044 0.053 0.063 0.052 0.049 0.051 0.011 Table 2: Experimental results for linear ELP Table 3: % Difference in optimal costs for the capacity computed using Large Deviation Bound vs. optimal costs for the capacity computed using exact formula for the exponential service and using a queuing simulator for the non-exponential service.

Table 4(a): Properties of the optimal solution (averages) for different service time distributions and waiting time bounds Table 4(b): Properties of the optimal solution (averages) for cost ratios and coverage radii Table 5: EFC properties of the optimal location vector. For the cases where the optimal vector is not EFC, “EFC Gap” measures the excess “structural capacity” in the system