Replica Dissemination and Retrieval Protocol Overhead

We evaluated the REDMAN protocols for replica distribution and retrieval in stationary scenarios. Mainly two tunable parameters have demonstrated to affect the SID overhead value: the number i of nodes hosting IRPs (that is the number of nodes along the straight path where replicas are positioned) and the maximum number s of hops explored in the retrieval phase. We ran more than 1,000 simulations with i and s independently varying from 2 to 15 hops (in the case distribution/retrieval paths reach network boundaries before arriving at their maximum number of allowed hops, REDMAN automatically makes paths continuing with a new random direction back in the dense MANET). Each plotted value represents the average of 20 simulations where delegates, dynamically determined by REDMAN RD, distribute i IRPs, and randomly chosen clients look for a resource replica by exploiting an s-hop-limited query.

Figure 1 shows the number of messages sent to replicate a resource at increasing distances, thus diffusing a higher number of IRPs. The plot shows that the message number linearly increases with i. Even if we choose high values of parameter i, the number of replication messages still remains largely acceptable (with respect to the other exhaustive strategies).

Figure 1: Number of replication messages sent by SID while varying i

Figures 2 and 3 show two different perspective of the same results.

Figure 2 shows how the number of sent messages per search increases with respect to the search distance s and the replication distance i. As we could expect, when SID diffuses a higher number of IRPs (high values of the parameter i), the number of messages required for search drastically decreases. On the other hand, if the number of search hops (i.e., the parameter s) increases, the sent messages consequently increase.

Figure 2: Number of sent messages per search while varying i and s

Figure 3 plots the same results from a different perspective. The figure highlights how the number of messages required for each search decreases with the diffusion of a larger number of IRPs (the parameter i specified on the x-axis). On the other hand, if we increase the search distance s, the number of search message increases.

Figure 3: Number of sent messages per search while varying i and s