Overview of Fault Tolerant Techniques in Underwater Sensor Networks

In order to obtain relevant sample in field of fault tolerant USN-s IEEE Explore, Google Scholar, Sciencedirect and Espacenet online environments were used with following search keywords in different combinations: “underwater”, “sensor network”, “internet of things”, “resilient”, “tolerant”, “fault management”, “cross-layer”. Top papers were selected by relevance order offered by environments’ algorithms and sources on the topic found from those papers. Other citations of those sources were searched and more papers found this way. Related articles offered by IEEE Explore and Sciencedirect algorithms were also taken into account. Collected papers were next analyzed, classified and divided into marine and terrestrial categories, and the number of papers managing specific areas of research are shown on radar diagram Fig. 2. It should be noted that in the context of Fig. 2 meaning of “localization” is location detection in space and meaning of “mobile” is capacity of movement. It can be seen from Fig. 2 that large share of marine research interest from the found papers has been drawn to underwater wireless communication while some on underwater fault tolerance techniques and almost none to underwater cross-layer fault tolerance. Underwater energy-efficiency and scalability are more covered areas than underwater vehicles (mobility) and security. Terrestrial papers were, according to initial search criteria, more concerned on fault-tolerance, including cross-layer fault tolerance, and less on energy-efficiency or security.

High research effort on marine wireless networking in Fig. 2 conforms the claim that current pace of research on Internet of Underwater Things is slow due to the challenges arising from the uniqueness of underwater wireless sensor networks  [1]. Specifically, the main challenges for Internet of Underwater Things are the differences between Underwater Wireless Sensor Networks and Terrestrial Wireless Sensor Networks  [1].

A fault is defined [2] as an underlying defect of a system that leads to an error. Error is a faulty system state, which may lead to failure, and failure is an error that affects system functionality. Faults may occur in different components and layers of systems for different reasons. The only type of fault possible in software is a design fault introduced during the software development i.e bug [3]. The software bugs can be addressed separately and will not be covered further in current paper.

Fault sources can be categorized by components where they occur. In sensor networks they can occur in sensor nodes, network and the data sink [4]. Sensor networks share common failure issues with traditional networks as well as introduce node failures as new fault sources [5].

USN-s introduce additionally faults caused by environment conditions such as pressure, currents, underwater obstacles, etc. Those conditions may cause physical damage that may result in failures as well as obstruct system functionality. For instance in underwater acoustic networks loss of connection and high bit error rate may be caused by shadow zones [6] formed by different physical reasons. Domingo and Vuran (2012) [7] distinguish up to 5 different underwater propagation phenomena which may obstruct communication.

Faults can either be permanent or temporary [8]. Permanent faults may be caused by manufacturing defect, as variances of the hardware components are inevitable due to physical reasons [9]. One of the other factors that can introduce faults is aging and wear-out of the hardware components [10]. In addition to the components themselves also the interconnection between them are affecting reliability and may cause faults [11]. One of the challenges of fault management is temporary faults, especially soft errors. Soft Error is a temporary change of signal value due to ionizing particles [8] that may lead to failure. Due to high integration density it is estimated that soft failure rate is increasing in future [12].

A distributed system is defined [13] as a collection of independent computers that appears to its users as a single coherent system. A Sensor network consists of a number of sensor nodes that form the network and feed data to single or multiple data sinks. Provided that in the sensor network the sensor nodes are autonomous, it can be seen as a distributed system. Faults happening in sensor networks can be addressed using the same techniques as in distributed systems [13]. The used techniques can be categorized into following groups:

• •

  Fault Prediction and Prevention are about preventing a fault to happen and proactive fault avoidance.

• •

  Fault Detection and Identification are responsible for detecting and localizing of the fault.

• •

  Fault Isolation, Masking and Recovery are different techniques for repairing fault, minimizing the effect of a fault or avoiding it to turn to system failure.

In Wireless Sensor Networks (WSN), instead of a centralized homogeneous topology, dividing nodes into clusters is an energy efficient and resilient method [14], where dedicated cluster head nodes may have more energy and communication capabilities to effectively act as mediators between regular nodes and data sinks.

To overcome issues caused by varying environmental challenges of Underwater Wireless Sensor Networks (UWSN), natural algorithms may be utilized. For instance clustering and routing can be done utilizing Cuckoo Search algorithm and Particle Swarm Optimization [15] which have behaved more resiliently in underwater conditions than more usual terrestrial Low Energy Adaptive Clustering Hierarchy (LEACH) protocol [16]. Pressure measurements have been used for UWSN routing [17] with floating depth-controlling sensors. Fault Management tasks can also be distributed across the whole network. In WSN with enough spare nodes energy efficient grid can be formed [18], changing node manager, gateway and sensing nodes selected and spare nodes put to sleep. This results in energy-efficient and lightweight network but needs excess nodes.

However, existing UWSN protocols have not been adequately compared in underwater field trial yet [19].

Sensor network deployment techniques are important for WSN where deployment may affect directly nodes’ locations and networking availability. Even for terrestrial wireless sensor networks, to obtain a satisfactory network performance, an adaptable deployment method is essential [20]. Usually sensor placement for WSN utilizes more sensors than the minimum required number for redundancy reasons [21]. The deployment costs and energy efficiency of WSN-s have been investigated and it has been found that there is no single solution that can be easily applied in practice [22].

Wired sensor network deployment is less researched, possibly because wired sensor networks’ node deployment locations are limited to cable, their locations are more predetermined and node connectivity is not directly related to location.

Sensor networks tend to have limited network bandwidth, energy and storage capabilities, thus filtering and aggregating sensor information may be a way to meet requirements. Raw sensor data near the source can be divided into informative, non-informative and outlier groups [23] and only needed data communicated or stored. Outlier data may result from noise, failures, disturbances etc. and may be useful for fault tolerance purposes.

Different techniques to compress and aggregate collected information in UWSN-s are investigated [24]. I was found that aggregation is justified and cluster-based aggregation techniques are performing better than non-cluster based or other.

Security challenges need to be addressed and one way to minimize the risk of data tampering and/or interference is to ensure that data is processed locally or if that is not possible then communicated end-to-end encrypted [25].

For UWSN-s there have been developed DESERT framework version 1 and version 2 [26] and SUNSET framework [27] that allow simulation, emulation and testing of networks. A conducted analysis [28] shows that SUNSET represents a more mature, flexible and robust framework for in field testing than DESERT, but DESERT v2 was released after that. For acoustic UWSN security testing SecFUN framework [29] has been proposed.

Faults and security are interrelated concepts [41]. It needs effort to prevent systems from being penetrated when working as intended, faults add uncertainty and make the task of prevention even harder. Faults can be created by an intrusion, but moreover faults can enable new intrusion vectors [44] - misbehaving devices violate key assumptions and create number of new attack vectors to systems. For example soft errors explained in section I-B can be used to defeat cryptography [45].

Power dissipation has now reached a point where energy concerns limit the computation we can deploy on chip [44] and the aim is shifting from transistor density and speed to energy density and cost. Energy density and efficiency needs to be addressed also on larger-scale, for instance WSN-s may not have unlimited power supply and need to utilize energy-efficiency strategies [22, 18, 14, 16]. For fault tolerance techniques, cross-layer approach is considered more energy-efficient  [30] than single layer. Strategic redundancy in cross-layer approach may allow systems to safely operate on the verge of failure [44] spending less energy without going over the edge.

One of traditional benefits of scaling has been the decrease of cost per functionality [44], but easing reliability problems by multiplicating logic, voting and similar techniques means that the scaled technology might not offer a reduction of energy or area. Some tolerance techniques may increase computing overhead, and not all approaches are scalable [42]. Large scale fault tolerant systems are researched [41] without paying special attention to energy and communication constraints.

Faults are not going to disappear but likely to increase in future [12]. One way to cope with faults is to accept imperfect devices to fail and compensate failures at higher levels in system stack [44], tolerating faults cross layer involving circuit design, firmware, operating system, applications etc. Cross-layer fault tolerant systems have potential to implement reliable, high-performance and energy-efficient solutions without overwhelming cost [30] by distributing the responsibilities of tolerating faults across multiple layers [46].

In case fault detection and fault recovery are to be implemented in different system layers then following challenges arise [47]:

• •

  For statistical validation and metrics high confidence resource-light reliability and availability estimation is needed.

• •

  Verification of resilience techniques, to be sure that resilience techniques perform under all possible scenarios.

• •

  Reliability grades for testing and grading system-wide reliability and data integrity. Reliability may change under workload.

In addition to Cross Layer approach also Multi Layer approach [48] has been proposed, where system layers are adapted to each other to reduce error propagation. In opinion of the author of current paper, this is not a separate approach, but rather a small increment of Cross Layer approach.

Underwater environment is mostly different because of harsh physical conditions - pressure, hard accessibility, limited communication and energy resources. Many communication methods are unavailable underwater and there are multiple phenomena [6, 7] that obstruct communication. Because of the possibility of flooding hardware, more attention and resources should be paid to physical security. On the other hand faults from excessive heat should be rare and avoidable underwater.

While most of the common concepts should be possible to be adapted for underwater use, the environment is more demanding and unforgiving and faults are more costly. Some more demanding approaches like cloud computing may not make sense to implement in USN, but author cannot see any low network bandwidth and power requirement fault tolerant approach mentioned in current paper, that cannot be used underwater. One of the more promising approaches that could be adapted well between underwater environment’s constraints seems to be cross-layer resilience, which for unknown reasons is lacking recent research papers even for terrestrial implementations.