15.4. Current Survivability Approaches in Wireless Networks

Survivability of wireless networks has recently begun to receive attention, mainly focusing on database survivability tailored to the cellular network databases (HLR and VLR) architecture [17–19]. This work focuses on the development of checkpoint algorithms and authentication techniques for the fault recovery of database contents. Literature on the design of a survivable landline topology for wireless networks [20–23] concentrates on formulating various optimization models for single-link failure survivable landline mesh-topology and capacity allocation design. However, the approach and assumptions used are identical to techniques used for wired backbone network design. None of the unique aspects of wireless networks are incorporated into the models. In Snow et al. [24], the importance of survivability for wireless networks is discussed and the difficulty in applying standard wired metrics for quantifying wireless survivability is shown.

We believe that survivability approaches for wired networks are not entirely applicable to the mobile and wireless domain. Consider that the failure of nodes and links is a primary survivability consideration. Wired networks are characterized by relatively high-speed, highly reliable, fixed-capacity links serving fixed users. The number of physical cables and their interconnection configuration influences system capacity. Diversity techniques often consist of adding spare capacity by the addition of physical cables, which are primarily subject to cost constraints. In contrast, the wireless domain is characterized by variable capacity and unreliable links serving mobile users. Wireless link capacity is influenced by the continually changing network conditions such as cell congestion, environmental factors, and interference. Diversity techniques are constrained by cost and a regulated frequency spectrum. Spectrum is a scarce resource in wireless networks and allocating spare capacity is much more difficult since, unlike wired networks, duplicating the medium is not easy. Furthermore, wireless network survivability approaches must account for user mobility and radio resource management. It has been shown that user mobility worsens transient conditions as disconnected users move among geographical areas to attempt to reconnect to the wireless access network [25–27]. In [25–27], the results of a sample survivability analysis of a typical GSM cellular network are presented using simulation. The steady-state and transient behavior of standard performance metrics, such as the call-blocking rate and location registration delay, were measured for a variety of failure cases. One significant result from these studies is that the impact of failure is larger than the failed area. For example, the failure of a BSC knocking out a group of four adjacent cells (in a network of 100) results in the mean time to process a location update for the entire group of 100 cells to exceed (by a factor of 10) the recommended International Telecommunications Union (ITU) benchmark value, resulting in protocol timeouts. Further, the magnitude and duration of a failure impact depend on a complex set of factors including the location of the failure (e.g., center or edge of location area), shape of the failed area (e.g., adjacent or disjoint cells), user mobility patterns, and user behavior in attempting reconnection. Thus, the network design should consider transient conditions in the capacity allocation and restoration techniques must consider spatial and temporal properties and address both the transient and steady-state periods. Current literature also does not consider the impact a failure in the wireless access networks has on the signaling network. In fact, our studies [26, 27] show that radio-level failure (e.g., loss of a base station, BS) causes a large increase in transient congestion in the signaling network. Similar results from network measurement on a GSM after an earthquake have been reported in the IST Caution project [28]. Incorporating such transient effects into the network topology design was recently proposed in [29].

A framework for cellular network survivability at the access layer, transport layer, and intelligent layer is presented in Tipper et al. [4, 25] along with potential survivability approaches at each layer. An extension of this framework to packet based 3G cellular networks was presented in [30]. Note that hybrid wireless networks offer alternatives to restoration schemes by enabling the use of overlays or underlays after failure of components of one particular technology. However, such schemes have also not been studied extensively for their benefits.

An overview of survivability issues and possible approaches for mobile ad hoc networks was presented in [31], including store-and-forward when end-to-end paths do not exist and store-and-haul in which nodes physically transport data. More recently, work on delay tolerant networking [32] has developed an architecture that has lead to work in disruption-tolerant networking. Disruption tolerant networking addresses the problem of how to communicate when wireless channel environments and mobility are so severe that stable end-to-end paths never exist and conventional communication that assumes routing convergence breaks down. In sensor networks, partitioning of the network due to failed sensor nodes (either hardware failure or battery exhaustion) can lead to the sensor network being unable to sense data or transmit the sensed data reliably to the base station. Work has been done to reduce the energy consumption in WSNs and load balancing among sensor nodes, but this is primarily at the academic research level.