DisService
DisService is an information dissemination middleware designed for tactical network environments. Tactical networks, the basis for Network Centric Operations, provide one of the most challenging environments for communications, with mobile nodes connected via limited bandwidth and highly variable latency wireless ad-hoc links in hostile RF environments. The dynamic nature of military operations further complicates the scenario, leading to a frequently changing network topology and widely varying loads being placed on the network by users and applications. Information dissemination, a critical function to enable essential tactical applications such as Blue Force Tracking, sensor data acquisition, target designation, and remote monitoring, is extremely difficult in this environment.
Fig. 1. The Tactical Edge Network Communication Environment
DisService was designed to opportunistically discover and exploits excess communications, storage, and processing capacity in a distributed network to improve the performance of information dissemination. It supports the storage and forwarding of data and caches data throughout the network, thereby making it disruption tolerant and improving the availability of data. Information is published in the context of a group, and may also be tagged to differentiate between multiple types of data (e.g., blue-force tracking, sensor data, logistics, or other runtime information). Each node in the network running DisService operates in a distributed, peer-to-peer manner while processing and communicating the published information and requested subscriptions from neighboring nodes. DisService disseminates information using an efficient combination of push and pull mechanisms, depending on the number of subscribers, the capacity of the network, and the stability of nodes in the network. DisService takes into account the user's preferences in order to anticipate information requests as well. In addition, it proactively fragments large data objects and replicates the fragments in order to improve sharing and increase availability of large data objects.
Opportunistic dissemination based on mobility prediction
DisService can adapt the dissemination strategies according to the current network conditions in order to achieve significant performance and reliability improvements. For instance, information that must be delivered to a small set of nodes that get periodically disconnected from the rest of the network might be best handled by dissemination strategies with aggressive caching and conservative forwarding. DisService integrates an adaptive system that chooses and tunes the dissemination strategies to adopt according to the current environmental conditions.
In this context, a particularly interesting metric to analyze is the contact window between different nodes. We define a contact window between nodes N1 and N2 as the tuple containing the start time and the duration of a time interval in which N1 and N2 are in communication range and can be considered neighbors. By monitoring and analyzing contact window information, DisService can discover cyclic mobility patterns and predict when a certain node might be in range.
We set up a realistic tactical edge network simulation environment with ns3 and studied contact window data. We observed that contacts between nodes exhibit an interesting behavior. In fact, the Fig. 2 shows the density estimation of contact window duration for each type of neighbor node, in order to verify which kind of node exhibits predictable contact window duration patterns.
Notice that the contact window distribution is multimodal. These results suggest the opportunity to predict the duration of contact windows by leveraging previously sampled values. In some cases, such as contact windows to the TOC, conditional prediction seems particularly interesting: when a certain threshold is passed, there is a high probability that the contact will last significantly longer. On the contrary the contacts to the UAV are mostly bounded by a 5-second threshold.
Fig. 2. Density Estimation of Contact Duration
In addition, the next figure shows the density estimation of time to next contact, grouped by neighbor node type. Fig. 3 demonstrates that time to next contact is a predictable metric. For instance, times to the contact with the TOC are mostly grouped around the main mode, at 22 seconds.
Fig. 3. Density Estimation of Contact Start Time
These observations raise the opportunity to study whether we are able to predict the next contact window information of a specific node with the algorithms described in the previous section. In addition, we are interested in finding which algorithm provides the most accurate results, and can effectively provide reliable input for dissemination strategy selection and tuning decisions.
Our findings [1] demonstrate that mobility prediction based on contact window analysis allows the development of opportunistic epidemic dissemination strategies which significantly improve the reliability of the system while limiting the retransmission of packets, since the messages are forwarded only when the destination can be reached. More specifically, in our testbed we have observed an average reduction of about 3% of the
network overhead, measured as the number of duplicate messages received by each node. The reliability of the system, instead, measured as the number of messages received from the UAV, significantly raised from an average of 75%-80% to almost 100%.
Publications
[1] A. Mazzini, C. Stefanelli, M. Tortonesi, G. Benincasa, N. Suri, "DisService: Network State Monitoring and Prediction for Opportunistic Information Dissemination in Tactical Networks", accepted for publication in Proceedings of IEEE MILCOM 2010 Military Communications Conference, October 31-November 2010, San Jose, CA, USA.
[2] S. Rota, G. Benincasa, M. Interlandi, N. Suri, B. Bonnlander, M. Tortonesi, J. Bradshaw, "Supporting Information on Demand with the DisServicePro Proactive Peer-to-peer Information Dissemination System", accepted for publication in Proceedings of IEEE MILCOM 2010 Military Communications Conference, October 31-November 2010, San Jose, CA, USA.
In collaboration with:
