Total order in opportunistic networks

Abstract

Summary Opportunistic network applications are usually assumed to work only with unordered immutable messages, like photos, videos, or music files, while applications that depend on ordered or mutable messages, like chat or shared contents editing applications, are ignored. In this paper, we examine how total ordering can be achieved in an opportunistic network. By leveraging on existing dissemination and causal order algorithms, we propose a commutative replicated data type algorithm on the basis of Logoot for achieving total order without using tombstones in opportunistic networks where message delivery is not guaranteed by the routing layer. Our algorithm is designed to use the nature of the opportunistic network to reduce the metadata size compared to the original Logoot, and even to achieve in some cases higher hit rates compared to the dissemination algorithms when no order is enforced. Finally, we present the results of the experiments for the new algorithm by using an opportunistic network emulator, mobility traces, and Wikipedia pages.

Supporting conflict-free replicated data types in opportunistic networks

“…Yet implementing and using CRDTs in OppNets has not been addressed much in the literature so far. Costea et al formulate two propositions, [10] and [11], in which a CRDT derived from Logoot [34] is used as a means to obtain total order in OppNets. In both of these papers the authors implement operation-based CRDTs, assuming that an opportunistic communication layer is available to ensure the causal dissemination of operations network-wide.…”

“…In both of these papers the authors implement operation-based CRDTs, assuming that an opportunistic communication layer is available to ensure the causal dissemination of operations network-wide. Two protocols are considered to ensure this dissemination (pure epidemic [33], and socially-aware interest-based dissemination [8]), and causal order is obtained based either on CBCAST (Causal Broadcast) [10] or on causal barriers [11]. The experimental results presented in both of these papers show that implementing operation-based CRDTs in OppNets puts significant pressure on the caches maintained on each node to store messages.…”

Causal and Δ-causal broadcast in opportunistic networks

“…To the best of our knowledge, causal broadcast in OppNets has so far only been addressed in [22] and [23]. In [22] causal broadcast is obtained with an algorithm that is derived from the CBCAST algorithm [15], and in [23] an algorithm derived from [11] is used instead. In both papers, though, the focus is on supporting conflict-free replicated data types (CRDTs) in OppNets, so very little detail is provided about how causal broadcast is actually implemented.…”