Furthermore, in VQB, the length of a virtual queue is used for making scheduling decisions.
Based on Lemma 1, we can conclude the stability of both the real and virtual queues and thus the stability of VQB, i.e., its throughput optimality.
Next, we discuss the overhead of VQB. The only additional overhead introduced by VQB is that used for maintaining per-flow virtual queues at each node.
In this section, we evaluate the performance of the VQB algorithm through simulation results.
In the simulations, we compare the VQB algorithm with the BP algorithm.
3 compares the BP algorithm with the VQB algorithm for the diamond network topology shown in Fig.
3(b) shows the overall packet delivery time with the BP algorithm and the VQB algorithm, respectively.
3(c) compares the average sum of the queue lengths of all nodes per timeslot with the BP algorithm and the VQB algorithm, respectively.
Finally, it is important to indicate that when all nodes in the diamond network have similar cache sizes (e.g., in case 1 where [RVQ.sub.(B)] = 1), VQB performs similarly to BP, which can be seen in Figs.
4 compares the BP algorithm and the VQB algorithm in terms of the packet delivery ratio, the overall packet delivery time, and average queue lengths for the grid network shown in Fig.