Maximizing data preservation in intermittently connected sensor networks

Proceedings of the 2015 Workshop on Mobile Big Data

Taase

2015

Self Cite

Many data-intensive sensor network applications are potential big-data enabler: they are deployed in challenging environments to collect large volume of data for a long period of time. However, in the challenging environments, it is not possible to deploy base stations in or near the sensor field to collect sensory data. Therefore, the overflow data of the source nodes is first offloaded to other nodes inside the network, and is then collected when uploading opportunities become available. We call this process data preservation in sensor networks. In this paper, we take into account spatial correlation that exist in sensory data, and study how to minimize the total energy consumption in data preservation. We call this problem data preservation problem with data correlation. We show that with proper transformation, this problem is equivalent to minimum cost flow problem, which can be solved optimally and efficiently. Via simulations, we show that it outperforms an efficient greedy algorithm.

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Data Preservation in Data-Intensive Sensor Networks With Spatial Correlation

Crary

Proceedings of the 2015 Workshop on Mobile Big Data

Taase

2015

Self Cite

“…However, most of the techniques focus on reducing the access cost [8,28,29,35] or battery consumption [24], and are therefore not suitable towards achieving K-Availability. Only line of research that focuses on enhancing data availability in sensor networks is called data redistribution or data preservation [10,27,30,31,34], which address how to preserve data inside sensor networks by tackling either storage-or energy-depletion induced data loss. Tang et al [30,31] study energy-efficient data redistribution in sensor networks wherein data is moved from nodes with highly loaded storage space to nodes with surplus storage, while minimizing the total energy consumption.…”

Section: Related Workmentioning

confidence: 99%

“…Overcoming the obstacle of data loss and preserving data until next uploading opportunity is therefore an important problem. In our previous research, we have addressed storage and energy depletion induced data loss [10,27,30,31,34]. However, unlike storage and energy depletion, sensor node hardware failure is unpredictable and thus can not be alleviated directly using similar techniques.…”

mentioning

confidence: 99%

Achieving data K-Availability in intermittently connected sensor networks

2014 23rd International Conference on Computer Communication and Networks (ICCCN)

Jaggi

Takahashi

2014

Self Cite

We consider the problem of preserving data in intermittently connected sensor networks wherein sensor nodes do not always have connected paths to the base stations. The generated data is first stored inside the network before being uploaded to the base station when uploading opportunities arise. Each node has both limited energy level and limited storage space, and is associated with a probability of failure. To guarantee that at any moment, each generated data item is available for being uploaded in the presence of node failure, we propose to replicate K copies of each data item in the network, where K depends upon the node failure probability. We refer to the problem as Data K-Availability Problem (DKAP). DKAP is naturally divided into two phases:K-Availability creation and K-Availability maintenance. For K-Availability creation, we show that the problem is NP-hard for arbitrary data sizes, and that it is equivalent to minimum cost flow problem for unit data sizes. For K-Availability maintenance, we show that it is NP-hard even for unit data sizes and design a centralized greedy heuristic. We further design an efficient and low-overhead distributed algorithm, which is applicable to both phases, and show using extensive simulations that it performs close to the heuristic.

“…The solution to avoid such data loss is simple: the overflow data is offloaded to other nodes with available storages (referred to as storage nodes). 1 Different data offloading techniques have been proposed with the goals of either minimizing the total energy consumption during data offloading [17], or maximizing the minimum remaining energy of storage nodes to prolong network lifetime [8], or offloading the most useful information considering data could have different priorities [21]. However, these techniques did not address the second level of data overflow, which is overall storage overflow explained below.…”

Section: Introductionmentioning

confidence: 99%

DAO-R: Integrating Data Aggregation and Offloading in Sensor Networks via Data Replication

Alhakami

2015 IEEE Global Communications Conference (GLOBECOM)

Han

et al. 2014

Self Cite

We study overall storage overflow problem in sensor networks, wherein data-collecting base station is not available while more data is generated than available storage spaces in the entire network. Existing research designs a two-stage solution to solve this problem. It first aggregates overflow data to the size that can be accommodated by the available storage capacity in the network, and then offloads the aggregated data into the network to be stored. We refer to this naive twostage solution as DAO-N. In this paper, we demonstrate that this approach does not necessarily achieve good performance. We propose a more unified method that is based upon data replication techniques, referred to as DAO-R, in order to improve the performance of DAO-N. Specifically, we design two energyefficient data replication algorithms to integrate data aggregation and data offloading in DAO-N. We show via extensive simulations that DAO-R outperforms DAO-N by around 30% in terms of energy consumption under different network parameters.