Continuous Answering Holistic Queries Over Sensor Networks

Continuous Answering Holistic Queries over Sensor Networks

Continuous Answering Holistic Queries overSensor Networks

ABSTRACT:

Sensor networks are widely used in various domains like the intelligent transportation systems. Users issue queries tosensors and collect sensing data. Due to the low quality sensing devices or random link failures, sensor data are often noisy. In order toincrease the reliability of the query results, continuous queries are often employed. In this work we focus on continuous holistic querieslike Median. Existing approaches are mainly designed for non-holistic queries like Average. However, it is not trivial to answer holistic onesdue to their non-decomposable property.We first propose two schemes based on the data correlation between different rounds, with onefor getting the exact answers and the other one for deriving the approximate results.We then combine the two proposed schemes into ahybrid approach, which is adaptive to the data changing speed.Weevaluate this design through extensive simulations. The resultsshow that our approach significantly reduces the traffic cost compared with previous works while maintaining the same accuracy.

EXISTING SYSTEM:

L-PEDAPs focusedon routing tree construction and maintenance for queryprocessing. Uncertainties may exist in both sensing dataand queries. For example, Zhang et al. studied theproblem of calculating the aggregate while the querylocation is uncertain.

Ye et al. proposed to determinepossible query results among all imprecise sensingdata.

Yu et al. presented a method which aiming at secure continuous aggregation querying.

DISADVANTAGES OF EXISTING SYSTEM:

It makes more sense to get a median result of the monitoring area than to derive an average result, since the noise may affect the average result largely

Due to the limitation of message size, the merging process may lose some information.

Generally, these methods can only achieve approximate results with some error guarantee by introducing different restricts and pruningalgorithms on the data structure

PROPOSED SYSTEM:

In this paper, we explore the correlation between data of different rounds and present two approaches to monitor continuous holistic queries, one for getting the exact answers and the other one for deriving the approximate results.

Then, we propose an effective hybrid approach that adaptively selects appropriate one based on data changing speed and improves the performance of continuous queries.

We propose a hybrid approach, combining F-Bucket and wavelet-like approaches, to handle the continuous holistic queries. If the sensor values in network are relatively stable, we apply an exact algorithm to calculate the exact median. When the data changes quickly, our approach can adaptively switch to the approximate algorithm.

We propose a histogram-based approach to get exact answers for continuous queries. Specifically, we use the histogram summary structure to store the value distribution of the network. Each bucket in the histogram counts the number of values in a certain range.

We propose two algorithms for refining the range assignment, Slip refining and Hierarchical refining.

ADVANTAGES OF PROPOSED SYSTEM:

This merging process reduces transmissions at the cost of losing some information. Finally, the sink aggregates all received AF-Buckets to an integrated one and calculates the query results.

We propose an effective hybrid approach that adaptively selects appropriate one based on data changing speed and improves the performance of continuous queries

We propose a histogram-based approach to get exact answers for continuous queries. Specifically, we use the histogram summary structure to store the value distribution of the network.

The metric number of transmissions to evaluate the traffic cost of all these approaches which is defined as the total size of transmitted data packets in one round.

SYSTEM ARCHITECTURE:

MODULES:

System Construction Model

Query Processing

Intermediate Node Operation

Result Calculation

MODULES DESCSRIPTION:

System Construction Model

In the first module, we deploy the system with the nodes/entities required to implement the proposed model. The nodes are deployed as the Sensor Nodes. We assume that a Median query is executed upon a datasetS within the rangeand the total number of sensorsin the network is n.in thispaper, we propose a histogram-based approach to getexact answers for continuous queries. Specifically, we usethe histogram summary structure to store the value distributionof the network. Each bucket in the histogramcounts the number of values in a certain range. The exactquery algorithm works when data values in sensor networkare relatively stable over time. Since the sensor valuesare relatively stable and highly correlated amongdifferent rounds, the Median result can be searched inmultiple rounds.

Query Processing

Although our query algorithm can work over arbitrarytopology structures, to simplify the discussion we assume atree-like routing topology. All sensor nodes in thenetwork are organized into a spanning tree rooted by a specialnode called sink. Take the Medianquery as an example, each round the leaf node builds an FBucketof only one bucket with its own value and transmitsto its parent. An intermediate node receives F-Buckets fromchildren and merges them with its own F-Bucket to an integratedone. The intermediate node then sends the new FBucketto its parent as well. In the end of the round, the sinkmerges all the received F-Buckets to a final one and calculates the median result. During a continuous Medianquery, the above process repeats round by round. The rangerefining algorithms finds the range where the median valueis located and subdivides it to all intermediate buckets.After a few rounds all the intermediate buckets handle arange of length 1 and the rest ranges are handled by twoboundary buckets. The median value is located in the rangeof intermediate buckets and exact median results can be calculatedcontinuously. When the median runs out the rangeof intermediate buckets, the range refining algorithm willadjust the ranges of buckets.

Intermediate Node Operation

The processing of an intermediate node is developed in this module. Basically, the intermediate node collects the FBuckets,merges them and sends the merged one to upperlevel. The merging of two F-Buckets is simple because therange of each bucket is fixed, we only need to merge theircount of corresponding buckets that are associated to thesame range.When sensor values change quickly, the range refining processwill be called frequently which leads to very high communicationcost.

Result Calculation

With the received F-Bucket in sink, we can answer a Medianquery. This module implements the steps to compute a Medianresult. According to the system, the return value is a rangethat covers the query result. If the returned range is oflength 1, we can get the exact median.At the very beginning, all buckets are equal length; duringthe continuous query, the range assignment will beadjusted to fit the value distribution better. The refiningprocess assigns more buckets to the range where the queriedmedian is located and adjust this assignment whilethe value distribution changes.it is clear that when thedata changing rate is low, the exact query schemeachieves very high efficiency. However, when sensor valueschange quickly, the range refining process will becalled frequently which leads to very high communicationcost. In this case, the approximate approach becomesmore stable to answer the queries. However, how toadaptively select the appropriate scheme is non-trivial.To fully leverage the advantages of both methods, in thiswork, we propose a novel metric named efficiency whichmodels how quickly the sensor value changes and itseffect on query processing. Using this metric, our hybridscheme adaptively applies different query algorithmsunder different circumstances during the query processing.

SYSTEM REQUIREMENTS:

HARDWARE REQUIREMENTS:

System: Pentium Dual Core.

Hard Disk : 120 GB.

Monitor: 15’’LED

Input Devices: Keyboard, Mouse

Ram: 1GB.

SOFTWARE REQUIREMENTS:

Operating system :Windows 7.

Coding Language:JAVA

Tool:Netbeans 7.2.1

Database:MYSQL

REFERENCE:

Kebin Liu, Member, IEEE, Lei Chen, Member, IEEE, Yunhao Liu, Fellow, IEEE,Wei Gong, Member, IEEE, and Amiya Nayak, Senior Member, IEEE, “Continuous Answering Holistic Queries overSensor Networks”, IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS, VOL. 27, NO. 2, FEBRUARY 2016.

Contact: 040-40274843, 9030211322

Email id: ,