Availability-aware content replication in disaster response networks

Abstract

In this thesis, we design and develop a content replication technique for replicating contents in disaster response networks. Disaster response networks evolve after a disaster when usual communication (e.g., the Internet) does not work due to damages or power outage. This kind of networks is a form of Delay-Tolerant Networks (DTNs) that leverage people's mobile phones, devices with rescue workers, volunteers and routers at vehicles or buildings to carry and forward messages around. Replicating important contents (such as phone books, maps of the rescue areas and various logistic information) for higher availability is an important issue in this context where popular replication services such as Dropbox and Google Drive, may not be available during that time. The only option prevails is to copy these contents to some other \nearby" nodes who are more \available" than others. We compute the availability of nodes by modeling the contact process among nodes by using a Markov chain and then analytically derive the expected retrieval time when a content object is replicated onto a set of replica nodes. Our objective is to choose this set so as to minimize the expected retrieval time from at least one such replica. We design a corresponding protocol, named availabilityaware content replication (ACR) protocol, that provides replication service in DTNs. We perform evaluation of our replication schemes in post-disaster mobility model (PDM) using the Opportunistic Network Environment (ONE) simulator. Our experimental results demonstrate that our proposed schemes perform better than other techniques such as randomized replication.