Fig. 1. Example of fragments A and B merging through the most central node.

Fig. 2. The finite state automaton for the dynamic case algorithm.

Fig. 3. Competitiveness distributions for Steiner heuristics (200-node networks) with 10% membership.

Fig. 4. Competitiveness distributions for Steiner heuristics (200-node networks) with 20% membership.

Fig. 5. D-ADH vs. D-ADHF comparison (a) D-ADHF solution cost percentage deviation from D-ADH and (b) D-ADHF round number percentage reduction from D-ADH.

Fig. 6. D-ADHF average time and number of rounds improvement from D-ADH.

Fig. 7. Cumulative percentage of networks solved (200-node synthetic networks with 10% membership) (a) within a given number of messages and (b) within a given time.

Fig. 8. Cumulative percentage of networks solved (200-node synthetic networks with 20% membership) (a) within a given number of messages and (b) within a given time.

Fig. 9. Cumulative percentage of networks solved (networks deployed on a cluster of real nodes) (a) within a given number of messages and (b) within a given time.

Fig. 10. Average convergence time (a) when the multicast group size is varied from 10% to 50% (100-node network) and (b) when the network size is varied from 50 nodes to 250 nodes (10% group membership).

Fig. 11. Average number of messages transmitted (a) when the multicast group size is varied from 10% to 50% (100-node network) and (b) when the network size is varied from 50 nodes to 250 nodes (10% group membership).

Fig. 12. Dynamic case analysis results (a) Competitiveness distributions for different heuristics and (b) Cumulative percentage of networks solved within a given time.

Legend of symbols adopted in the paper

Test networks used in cost competitiveness experiments