Performance of BGP Among Mobile Military Networks.

摘要

As tactical military networks deploy new IP-capable radios (e.g., JTRS), it is expected that network connectivity wiII increase both within and among Service and Coalition networking domains (e.g., Army, Navy, NATO). However, unmanaged use of this Increased connectivity may result in violation of some mission-critical operational limits. One mechanism known to (coarsely) manage large-scale IP networks is BGP routing policy. As such, this paper begins the study of BGP's applicability to manage an evolution of the GIG In which IP-based tactical radios proliferate (i.e., the Future GIG). To this end, this paper first presents a modification to BGP that allows for dynamic management of its peering sessions to accommodate network node mobility. Next, since it is known that BGP can have performance issues (e.g., slow convergence), this paper uses network emulation(caret)1 to perform a performance assessment of our modified BGP protocol. Specifically, for ten independent realizations of a mobile wireless networking model, our modified BGP protocol is evaluated with respect to its generated protocol overhead, its ability to develop valid routes to destinations (e.g., reachability), and its influence on network's outage events. Furthermore, our modified BGP protocol is compared to the OSPF and OSPF-MDR routing protocols for mobile networks with increasing number of nodes. Our results show that modified BGP's overhead growth is significantly higher than both OSPF variations, but the network's average reachability and median outage times are similar. To decrease the significant overhead seen for BGP, this paper also presents a second modification to the BGP protocol based on the use of a connected dominating set (CDS) backbone. Here, the emulation results show that the use of a CDS backbone can significantly decrease BGP's overhead with little impact on the network's average reachability and median duration of its outage events. Lastly, it was found the maximum network outage times for both modified BGP protocols were tens of seconds greater than that that experienced by either OSPF-based routing protocols. However, the occurrence rate of long network outage events (e.g., greater than 60 seconds) was also seen to be infrequent for networks whose average shortest paths were less than 3 hops.