Probabilistic location-free addressing in wireless networks
Introduction
Assigning routable addresses to nodes in multi-hop wireless networks such as sensor, ad hoc and mesh networks is challenging. Apart from the inherent instability of wireless links, node mobility is a prominent, complementary factor that prompts unpredictable topological changes. In addition to link dynamics, these further complicate the provision and maintenance of a reliable addressing and routing topology in dynamic networks. This is because, perceived changes of node locations, whether due to link and node dynamics, typically cause the addressing scheme to trigger a change of its assigned address. Such address changes are not desired for two main reasons. First, each change of a node address sparks updates of this address and subsequent address queries across the network, resulting in substantial and expensive traffic overhead. Second, packets routed towards the respective node might never reach it due to the outdated address and typical impracticality of recovering previous addressing topologies.
With regard to existing addressing approaches, the flat and distributed topology and dynamics of wireless networks require dedicated approaches. For example, while still used for higher-layer naming purposes, hierarchical addressing schemes such as IP cannot reflect changing, distributed topologies and require centralized control (i.e., DHCP) to carefully (re-)allocate addresses in case of network dynamics such as link failures. To incorporate the underlying topology, network characteristics, and notion of routing costs, geographic routing, e.g., GPSR (Karp and Kung, 2000) may be employed to achieve scalable routing in wireless networks. However, efficient geographic routing is dejectedly dependent upon careful (re-)configuration, may require modifications in the commodity hardware (i.e., to install GPS), and cannot accommodate large scale mobility.
Recently, the concept of location-free or logical coordinate addressing has received much attention. Logical addressing assigns routable addresses to nodes based on the underlying connectivity graph and their logical location with respect to neighboring nodes. Nodes can thereby autonomously determine their own addresses without centralized mediation. However, while meeting the requirements of multi-hop wireless addressing for routing, the problem of arbitrary address changes remains unsettled, especially in case of node mobility in ad hoc and mesh networks. We thus propose a probabilistic addressing scheme (PAD) that deviates from fixed, discrete addresses but allocates routable regions to nodes. In PAD, regions absorb network dynamics to keep node addresses valid while they remain within a region. Each region is defined by the statistical distribution of hop distances of the respective node to a set of landmarks in the network. Indeed, a node needs to update its address in the distributed network-wide address-database only when leaving its region.
This paper makes the following key contributions:
Address stability: Compared to other addressing and routing schemes, PAD requires 3–7 times fewer address updates in a global location directory. At the same time, it maintains a small amount of state and requires considerably less effort and complexity in its mechanisms and implementation. We show that such stable addressing can be achieved even by considering only very recent link conditions instead of pessimistically overhearing and estimating links over a time period in the order of minutes (or hours).
Address monotony: Once an address update occurs, the difference between the old and new location of a node is 3–12 times smaller for PAD than for comparable approaches. This implies that the changes in PAD's addresses are gradual, which helps routing success. Our evaluation shows that this phenomenon allows PAD to maintain more up-to-date yet stable node locations in the network.
Responsiveness: By decoupling addressing from routing and link estimation, PAD can respond rapidly to changes in link quality which existing routing algorithms naturally avoid. As a result, each data packet can be forwarded on a different path depending upon the very recent network conditions. Our comparative analysis on four testbeds shows that even a simple routing strategy over PAD reduces the number of transmissions by 26% in testing network conditions when compared with S4 (Mao et al., 2007) – state-of-the-art cluster based point-to-point routing in sensor networks.
Generality: Finally, we show that the utility of PAD is not limited to any specific class of wireless networks. Rather, it has a broader relevance in the wireless domain and it carries its superior performance characteristics across different classes of wireless networks such as IEEE 802.15.4 based sensor, and IEEE 802.11 based mesh networks. We demonstrate the generality of PAD using both simulations and testbed evaluation.
In Section 2, we give an overview over the background, concept and related approaches of location-free, probabilistic addressing. We present the distinct parts in our design of PAD for dynamic sensor, ad hoc and mesh networks in Section 3. Section 4 evaluates the characteristics of PAD IEEE 802.15.4 based sensor networks. Section 5 highlights the performance and benefits of PAD in a simulated and in a real-world IEEE 802.11 network. We conclude the paper in Section 6 and discuss future work, notably the combination of PAD addressing with existing multi-hop routing protocols.
Section snippets
Overview
There are a few basic ingredients of the location-free addressing domain that are essential to grasp the material of this paper. Therefore, in this section, we briefly revisit location-free addressing before summarizing the concept of probabilistic addressing and how it can be crafted into PAD to meet the challenges of multi-hop wireless networks. We conclude this by identifying the target network environments and highlighting the related research efforts.
PAD design
Approaches such as BVR and S4 attempt to filter out the variability in a node's coordinates, which is caused by network dynamics, to obtain a stable address. In contrast, PAD incorporates this variability into a node's address by encoding a limited history of the node's coordinates. The idea is to learn from the dynamics exposed by a node's coordinates and express them in the form of probabilities. The routing algorithm can then determine a node's coordinates by predicting its current location
Evaluating PAD in IEEE 802.15.4
Our evaluation of PAD in IEEE 802.15.4 based networks focuses on two aspects: (1) We need to choose an appropriate history size σ and error level ε between PAD coordinate distributions (cf. Section 3.1.3). (2) We need to thoroughly compare PAD with existing virtual coordinate based addressing approaches to observe potential benefits and drawbacks of our approach. PAD is implemented for OMNeT++ and TinyOS 2.1.0. It has been tested in both OMNeT++ and TOSSIM simulators, and on IEEE 802.15.4-based
Evaluating PAD in IEEE 802.11
Although PAD's design is independent of link layer properties, we still cannot make any assumptions regarding its feasibility and performance on a different link layer technology. Rather, we need to demonstrate this empirically for two main reasons: (1) IEEE 802.15.4 is very different from other link layer technologies such as IEEE 802.11: for example, it operates on a constant bit rate, supports very low data rates (few tens of Kbps compared to tens of Mbps in IEEE 802.11), and its design is
Conclusion and future work
We presented a robust and scalable addressing mechanism for multi-hop wireless networks that incorporates node mobility with negligible maintenance overhead. When compared with other addressing mechanisms, PAD increases the stability and reduces the magnitude of change in addresses. An adaptive routing strategy over PAD allows quick adaptation of the routing paths based on very recent link conditions. Our results from testbed environments demonstrate that even an unoptimized version of routing
References (37)
- et al.
Portable wireless-networking protocol evaluation
J Netw Comput Appl
(2013) - et al.
Link stability estimation based on link connectivity changes in mobile ad-hoc networks
J Netw Comput Appl
(2012) - Aguayo D, Bicket J, Biswas S, Judd G, Morris R. Link-level measurements from an 802.11b mesh network. In: SIGCOMM,...
- Alizai MH, Landsiedel O, Bitsch Link JA, Goetz S, Wehrle K. Bursty traffic over bursty links. In: SenSys'09, November...
- et al.
Exploiting the burstiness of intermediate-quality wireless links
Int J Distrib Sens Netw
(2012) - Alizai MH, Vaegs T, Landsiedel O, Goetz S, Link JAB, Wehrle K. Probabilistic addressing: stable addresses in unstable...
- Allen GW, Swieskowski P, Welsh M. Motelab: a wireless sensor network testbed. In: IPSN,...
- Biswas S, Morris R. Exor: opportunistic multi-hop routing for wireless networks. In: SIGCOMM,...
- et al.
Scalable logical coordinates framework for routing in wireless sensor networks
ACM Trans Sens Netw
(2006) - Chakeres I, Perkins C. Dynamic MANET on-demand (DYMO) routing. Internet Draft, March...
Energy efficient routing with delay guarantee for sensor networks
Wirel Netw
Dartdynamic address routing for scalable ad hoc and mesh networks
IEEE/ACM Trans Netw
Cited by (1)
Practical and self-configurable multihop wireless infrastructure: A functional perspective
2017, International Journal of Ad Hoc and Ubiquitous Computing