“…In a nutshell, each experimental data point is realized with 1000 requests following a Poisson process VNFFG request whose mean inter-arrival time is set to 25 s, and the duration (holding time, HT) is exponentially modeled varying its mean to 200, 250, 300, and 350 s. This provides different offered traffic loads (expressed in Er): 8, 10, 12, and 14. The requirements of each VNFFG requests are generated as follows: the number of VMs per DC is uniformly distributed between [1,5]; the IT resources (i.e., CPU, RAM and disk) are randomly chosen in the ranges of [1,4] cores, [1,6] GB and [4,10,20,40] GB, respectively; the demanded bandwidth is randomly selected among [10,40,100] Gb/s; and the latency (l) is in the range of [6,12] ms. Fig. 4 shows the percentage of the accepted reqs versus the offered traffic load in Erlangs (Er) when applying either MD or LLDC.…”