Home // International Journal On Advances in Systems and Measurements, volume 3, numbers 3 and 4, 2010 // View article

Prioritized Redundancy of Data Storage in Wireless Sensor Networks

Authors:
Cosmin Dini
Pascal Lorenz

Keywords: sensors; networks; storage; operations; data management; data sensor storage; data priority; optimized parameters; prediction models; prioritized redundancy

Abstract:
Wireless Sensor Networks have evolved into complex deployments, where their nodes can have a full network protocol stack, database systems, etc. The main rationed resource in such a deployment is energy. Having power usage tightly managed ensures a long operational life for the node in cases where replenishments (either via recharge or battery change) are difficult or impossible. The basic deployment of Wireless Sensor Networks consists of sensing nodes as well as a relay node (i.e., sink), which collects sensory data to be relayed via a reliable network. The sink node can become unreachable due to malfunction, scheduled uptime or, in the case of mobile sink nodes, due to being out of the sensor nodes’ reach. In addition, the sensor nodes may decide against relaying data for some period. In these cases, optimal use of sensor node memory space also becomes critical. In this article, we classify data types and establish a set of node level approaches that can be taken to make the most of limited data storage via a prioritized data reduction. Wireless Sensor Networks often operate in locations with limited access while relying on a restricted set of resources. This calls for careful management of local assets such as energy and storage. Storage is used to host data of various interest levels for subsequent relay to a base station or for queries through the network. The size of the data needs to be managed in view of data’s relevance. While there is a debate for complete or partial data extraction from sensors, having special data process functions and operation primitives proven useful for sensor OSs. To deal with robustness and reliability, data processing at the network/sensor level satisfies some of the reliability requirements, especially when communications are not operational. There are situations were data reduction is an alternative when storage is not longer available and data is aging, especially when some sensor links are not properly operational. Using predictions and optimized parameters to prioritize data reduction is a solution. While approaching the data reduction from the perspective of a single node is important, there are benefits from looking at Wireless Sensor Networks deployment as a whole. There is an opportunity to relocate some data to less active nodes and spare it from a reduction process. There are energy considerations to take into account in evaluating the potential benefit from spending some energy to relocate data. Three factors play a part in this: the source node, the receiver node, and the data itself. The source node needs to save enough energy to relay its data once the sink node becomes available. The receiver node needs to have space available, or at least have a lot of space occupied by low importance data. The data itself needs to be self-enclosed as far as parameters needed for importance computation. Another aspect that becomes feasible in a multi node environment is having redundant copies of data to protect against potential node failures. These copies have their own particularities, a notable one being that their importance function cannot depend on other data. This paper presents a methodology that enables the node to be useful by collecting data beyond the point where its data storage size would otherwise allow. We build upon primitive data reduction operations to construct a framework that can be used for a sensible data age-out scheme. The methodology enables various factors, both internal and external to the sensor, to influence the data aging process and the data reduction operations. Additionally, we define special heuristics for data reduction using a set of data processing primitives and special data parameters. We apply these heuristics via a methodology that enables various factors, both internal and external to the sensor, to influence the data aging process and the data reduction operations.

Pages: 163 to 185

Copyright: Copyright (c) to authors, 2010. Used with permission.

Publication date: April 6, 2011

Published in: journal

ISSN: 1942-261x