Noname manuscript No. (will be inserted by the editor) Routing for Disruption Tolerant Networks: Taxonomy and Design Thrasyvoulos Spyropoulos · Rao Naveed Bin Rais · Thierry Turletti · Katia Obraczka · Athanasios Vasilakos Received: date / Accepted: date Abstract Communication networks, whether they are it very difficult, if not impossible, to design a routing wired or wireless, have traditionally been assumed to be solution that fits all. connected at least most of the time. However, emerging In this paper, we first break up existing routing applications such as emergency response, special opera- strategies into a small number of common and tunable tions, smart environments, VANETs, etc. coupled with routing modules (e.g. message replication, coding, etc.), node heterogeneity and volatile links (e.g. due to wire- and then show how and when a given routing module less propagation phenomena and node mobility) will should be used, depending on the set of network charac- likely change the typical conditions under which net- teristics exhibited by the wireless application. We fur- works operate. In fact, in such scenarios, networks may ther attempt to create a taxonomy for intermittently be mostly disconnected, i.e., most of the time, end- connected networks. We try to identify generic network to-end paths connecting every node pair do not exist. characteristics that are relevant to the routing process To cope with frequent, long-lived disconnections, op- (e.g., network density, node heterogeneity, mobility pat- portunistic routing techniques have been proposed in terns) and dissect different “challenged” wireless net- which, at every hop, a node decides whether it should works or applications based on these characteristics. forward or store-and-carry a message. Despite a grow- Our goal is to identify a set of useful design guidelines ing number of such proposals, there still exists little that will enable one to choose an appropriate routing consensus on the most suitable routing algorithm(s) in protocol for the application or network in hand. Finally, this context. One of the reasons is the large diversity of to demonstrate the utility of our approach, we take up emerging wireless applications and networks exhibiting some case studies of challenged wireless networks, and such “episodic” connectivity. These networks often have validate some of our routing design principles using sim- very different characteristics and requirements, making ulations. Keywords DTN routing · Intermittent connectivity · Thrasyvoulos Spyropoulos ETH Zurich, Switzerland Disruption tolerance · Message replication · Network E-mail: spyropoulos.tik.ee.ethz.ch coding · Routing design guidelines Rao Naveed Bin Rais INRIA - University of Nice, Sophia Antipolis, France E-mail: [email protected] 1 Introduction Thierry Turletti INRIA, Sophia Antipolis, France E-mail: [email protected] Traditionally, communication networks, regardless of Katia Obraczka whether they are wired or wireless, have always been University of California at Santa Cruz, USA assumed to be connected almost all the time. Here, E-mail: [email protected] by connected networks, we mean that there exists at Athanasios Vasilakos least one end-to-end path between every pair of nodes University of Western Macedonia, Greece in the network most of the time. When partitions occur, E-mail: [email protected] they are considered transitory failures and core network 2 functions such as routing react to these failures by at- epidemic routing offers minimum delivery delay, it may tempting to find alternate paths. Even in wireless multi- be prohibitively expensive since it consumes consider- hop ad-hoc networks (e.g., MANETs), where links are able network resources due to the excessive amount of more volatile due to wireless channel impairments and message duplicates generated. mobility, partitions are still seen as exceptions and as- Our focus here is on opportunistic approaches to sumed infrequent and short-lived. DTN routing, i.e., where no contact information is known However, for some emerging applications like emer- a priori and no network infrastructure (e.g., special- gency response, special operations, smart environments, purpose nodes with controlled trajectories) exists to habitat monitoring, and VANETs, which are motivated provide connectivity. Besides the question of when con- by advances in wireless communications as well as ubiq- tact opportunities happen between nodes, a number uity of portable computing devices, the assumption of of other factors also affect data forwarding, including “universal connectivity” among all participating nodes available storage at peering nodes, contact duration, no longer holds. In fact, for some of those scenarios available bandwidth, message priority or expiration time, or applications, the network may be disconnected most etc. of the time; in more “extreme” cases, there may never An ever growing number of protocols addressing be an end-to-end path available between a source and these “opportunistic” DTN scenarios have been pro- a destination. Besides the application scenarios them- posed. However, it is not at all clear how existing solu- selves, other factors contributing to frequent, arbitrarily tions can be applied to a variety of DTN applications long-lived connectivity interruptions include node het- given their requirements and underlying network char- erogeneity (e.g., nodes with different radios, resources, acteristics (e.g., connectivity, node mobility and capa- battery life), volatile links (e.g., due to wireless propa- bility). gation phenomena, node mobility), energy efficient node In this paper, we address this question and thus help operation (e.g., duty cycling). map the design space of opportunistic DTN routing. We can summarize the contributions of this work as follows: Networked environments which operate under such intermittent connectivity are also referred to as episod- – First, we dissect opportunistic routing solutions iden- ically connected, delay tolerant, or disruption tolerant tifying their basic building blocks in terms of the for- networks (or DTNs). Clearly, traditional routing, in- warding scheme employed, namely message replica- cluding MANET routing protocols like OLSR [1], AODV tion, forwarding, and (source and network) coding [2], and DSDV [2] cannot deliver adequate performance (Section 2). in DTNs. Consequently, a number of new routing ap- – We also identify a number of features that can be proaches have been proposed to cope with frequent, ar- used to classify DTNs. Classifying DTNs accord- bitrarily long-lived connectivity disruptions. They can ing to their connectivity, mobility, and capability be classified into three categories: deterministic or sched- (i.e., storage, battery life, processing) of the partic- uled, enforced, and opportunistic routing. Deterministic ipating nodes will be key to deciding what routing routing solutions are used when contact information is mechanism(s) to use in order to achieve adequate known a priori. Jain et al. [3] showed how little or full application-level performance (Section 4). information about contacts, queues, and traffic can be – We then proceed to map the opportunistic rout- utilized to route messages from a source to a destina- ing design space by drawing the correspondence be- tion in the case of disruptions. They have presented a tween the proposed DTN taxonomy and the basic modified Dijkstra algorithm based upon information on opportunistic routing building blocks (Section 5). scheduled contacts and compare the proposed approach – Finally, through simulations, we conduct case stud- against an optimal LP formulation. In order to deliver ies of a number of challenged wireless network sce- messages to otherwise disconnected parts of network narios in order to validate some of our DTN oppor- (islands), enforced routing solutions like message ferries tunistic routing design principles and recommenda- [4] and data mules [5] can be employed, where special- tions (Section 6). purpose mobile devices move over predefined paths in order to provide connectivity. Epidemic dissemination The remainder of this paper is organized as fol- [6] is the basic form of opportunistic routing and works lows. Section 2 discusses the routing strategies in in- as follows. When node A encounters node B, it passes termittently connected network by dissecting the ex- to B replicas of messages A is carrying which B does isting solutions into a small number of common and not have. In other words, epidemic routing is to episod- tunable routing primitives. Important utility functions ically connected environments what flooding is to “tra- for routing decisions are described in Section 3. Sec- ditional”, well-connected networks. While on one hand tion 4 presents a DTN taxonomy by detailing the net- 3 work characteristics that are important in designing a ner. This is not the case in a DTN-like environment, routing protocol. DTN routing design guidelines and a as it is possible that a path may never be available be- discussion is presented in Section 5, and in the end, we tween source-destination pairs. Hence, the store-carry- provide some case studies of challenged wireless net- and-forward routing paradigm is utilized in such scenar- works in Section 6. ios; this means that a set of independent, opportunistic (i.e., no certainty about whether there will ever be a path to destination) forwarding decisions will attempt 2 Opportunistic Routing Primitives to eventually deliver messages