Opportunistic Networking : Congestion, Transfer Ordering and Resilience
Opportunistic networks are constructed by devices carried by people and vehicles. The devices use short range radio to communicate. Since the network is mobile and often sparse in terms of node contacts, nodes store messages in their buffers, carrying them, and forwarding them upon node encounters....
| Main Author: | |
|---|---|
| Format: | Doctoral Thesis |
| Language: | English |
| Published: |
Uppsala universitet, Avdelningen för datorteknik
2014
|
| Subjects: | |
| Online Access: | http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-223492 http://nbn-resolving.de/urn:isbn:978-91-554-8953-3 |
| Summary: | Opportunistic networks are constructed by devices carried by people and vehicles. The devices use short range radio to communicate. Since the network is mobile and often sparse in terms of node contacts, nodes store messages in their buffers, carrying them, and forwarding them upon node encounters. This form of communication leads to a set of challenging issues that we investigate: congestion, transfer ordering, and resilience. Congestion occurs in opportunistic networks when a node's buffers becomes full. To be able to receive new messages, old messages have to be evicted. We show that buffer eviction strategies based on replication statistics perform better than strategies that evict messages based on the content of the message. We show that transfer ordering has a significant impact on the dissemination of messages during time limited contacts. We find that transfer strategies satisfying global requests yield a higher delivery ratio but a longer delay for the most requested data compared to satisfying the neighboring node's requests. Finally, we assess the resilience of opportunistic networks by simulating different types of attacks. Instead of enumerating all possible attack combinations, which would lead to exhaustive evaluations, we introduce a method that use heuristics to approximate the extreme outcomes an attack can have. The method yields a lower and upper bound for the evaluated metric over the different realizations of the attack. We show that some types of attacks are harder to predict the outcome of and other attacks may vary in the impact of the attack due to the properties of the attack, the forwarding protocol, and the mobility pattern. === WISENET |
|---|