Volume 6 Number 1 June 2016


Time Slice based Advance Resource Reservation in Grid Computing Environment
S. Nirmala Devi ,A. Pethalakshmi

Abstract: Grid computing provides a promising environment for the accomplishment of a particular task by sharing the resources when required. Grid environment is highly dynamic and also heterogeneous. The resources which normally shared are processor, storage and network bandwidth. Discovering the resource which suits the requirement itself is a tedious task. Sometimes, even after finding the resource, it may not be available as it may be used by some other task. To get the resource at the required time the resource can be reserved in advance. There are various advance resource reservation schemes as FCFS, Alternate Offer Protocol, priority based reservation etc. In this paper a new reservation scheme called TARR Time-Slice based Advance Resource Reservation is proposed. In this scheme, the reservation is done when the resource is free. If the resource is already reserved during that timeslot then the free the time slices can be used for the reservation. This splits the resource utilization period, i.e., whenever a free time-slice is available the resource is reserved for that duration and the remaining is deferred over a period of time where the free time slice is available. By applying this approach the Average Waiting Time (AWT) of the job to be completed decreases, the Hit-Ratio increases for fetching the resources and even the Resource Idle Time (RIT) decreases.


A Comparative Study of Mean Value Analysis and Convolution Algorithm for Queueing Networks
Jitendra Kumar, Vikas Shinde

Abstract: Present paper deals, comparative study between mean value analysis and convolution algorithm for queueing network models by exponentially distributed service time for single and multi-class system. The throughput, utilization, average response time and mean number of jobs were obtained using performance measures. Numerical illustrations have been carried out to examine the effect of various parameters on performance indices.


Enhanced Hybrid Compression Methods For Compound Images
D.Banupriya, M.Sundaresan

Abstract: This work presents an efficient compound image compression method based on object, block and layer based segmentation techniques, which introduces a new hybrid scheme for segmenting compound images. Effective compound image compression algorithms require compound images to be first segmented into regions such as text, pictures and background to minimize the loss of visual quality of text during compression. This work discusses the relative advantages of each scheme and studies the use of fast classification techniques for segmentation that can be used together with chosen compression architecture. The algorithms have been developed and implemented to compress and decompress the given image using suitable techniques for each method in a MATLAB platform. The performance metrics like Compression ratio, PSNR, Compression and Decompression time are tested for six models.


IoT Data and its Application-A Preliminary Study
V. DiviyaPrabha, R. Rathipriya

Abstract: The world has been started moving from connecting things to capturing insights towards Internet of Things (IoT). Interconnection of devices through sensors continuously generates data 24x7 which leads to growth of data at exponential rate. Capturing, storing, processing and retrieving of these massive data is a tedious task and has a wide scope for research in the data storage and retrieval. Traditional database system does no satisfy the needs of IoT for maintaining data. Therefore, this paper suggests necessary things needed for IoT data management and its security.


Pest Identification in Leaf Images Using SVM Classifier
R.Uma Rani, Ms.Amsini

Abstract: This paper entitled ?Pest Identification in leaf images using SVM Classifier? is mainly developed to detect and calculate the accuracy of pest infected area in leaf images. In modern agricultural field, pest detection is a major role of plant cultivation. The production rate of crops is reduced in agricultural field by the presence of whitefly pests, aphids and thrips which cause leaf discoloration. The image segmentation technique is used to detect the presence of pests in leaf images. The performance of the clustering based image segmentation algorithm depends on its simplification of images. The K-means cluster algorithm has been proposed to identify the accurate location of whiteflies, aphids and thrips in various leaf images. The infected area is calculated by SVM classifier. The algorithm was developed and implemented using MATLAB 7.14 build 2012a.


Detection of Satellite Image Edges using B2MST
K P Sivagami, S K Jayanthi, S Aranganayagi

Abstract: In image processing and pattern recognition, edge detection is used to preserve the structural properties in an image which significantly reduces the amount of data and also filters out the useless information. Edge detection is an important area in processing the satellite images which are of high resolution with lot of information. Bi-level Bi-stage concept has been used to detect the edges based on global and local threshold values in Shannon entropy Multi thresholding (SMT) method for gray scale images. To extend this concept for multispectral images, a Shannon entropy Multispectral Multi Thresholding (SM2T) algorithm has been proposed. Even though this method detects more edges than SMT, Edge Detection using Multispectral thresholding (EDMST) method, based on Otsu thresholding values for multispectral images, detects more edges than SM2T method. Bi-level Bi-stage Multispectral thresholding (B2MST) algorithm has also been proposed based on global and local Otsu thresholding values to improve the EDMST method. Even though all the methods are applied on natural, art and simulated images, the performance is evaluated on simulated images, due to the existence of well-known edges. The result of SMT, SM2T, EDMST and B2MST methods have been compared based on human visual system, number of edges detected and F-measure. Finally it has been observed that the B2MST shows better results and hence applied on satellite images.


Optimized Algorithms for Virtual Machine Placement Based On Multi-Dimensional Resource Characteristics in Cloud Computing Systems
RT. Thiruvenkadam, P. Kamalakkannan

Abstract: Virtual machine placement to the PMs of the cloud datacenter is one of the important problems in cloud environment to provide better service to the cloud users. This research work proposed techniques to improve the performance of virtual machine placement in cloud environment. The proposed placement algorithm consisted of two main tasks. The first task optimizes the scheduling, while the second task enhances the operation of load balancing. The scheduling and load balancing is performed as a two-step process, where the first step groups resource requests into three categories, namely, high, low and medium resource requests queues and the second step performs scheduling and load balancing. The proposed VMP-LR, during non-rush hours, as the number of requests is minimum, uses an enhanced round robin method. During rush hours, in order to accommodate high resource requests, separate hybrid scheduling and load balancing algorithms are used to handle high, low and medium queues. The simulation results proved that the proposed algorithms are efficient in mapping VMs to PMs effectively in terms of cloud service response time and can save energy and increase resource utilization in a positive manner.


An Analysis of Space Query Classifier Indexing for Mining Uncertain Data
M. Kalavathi, P. Suresh

Abstract: Data uncertainty develops into an accepted topic in database and data mining area due to the extensive survival of uncertainty. The uncertain data is used in several real applications such as sensor network monitoring, object recognition, Location-Based Services (LBS), and moving object tracking. Due to the intrinsic property of uncertainty, many interesting queries are used for different purposes. Data uncertainty arises clearly and inherently in many applications. The causes of uncertainty in applications comprise data uncertainty, incompleteness, control of measuring equipment, the delay or loss of data updates and privacy preservation. Hence, this article mainly concentrates to solve the above mentioned tolerance problem and also reduces the overhead count.


Data Analytics Framework: R and Hadoop-Geo-location Based Opinion Mining of tweets
K. Santhiya, V. Bhuvaneswari

Abstract: Internet social media services such as Twitter have seen phenomenal growth as millions of users share opinions on different aspects of life every day. This tremendous growth has induced an interest in making use of such data for extracting valuable information, such as their opinions, location of the users and certain other information. In this paper we have analyzed the tweets related to crime attributes against women and children, different sort of crimes that are prevailing , the location in which the users tweets are more frequently occurring related to crimes. The proposed work make use of R language for extracting real time tweets and relies upon Hadoop-based framework for storing the tweets as they are larger in number. The tweets are parsed under Hive environment and we build a sentiment classifier in R that is able to determine positive, negative and neutral sentiments for a given phrase. We observe that the elapse time for processing under Hadoop based framework significantly outperforms the other conventional methods and is more suited for real time streaming tweets.


A Proficient Segmentation of Remote Sensing Images Using Modified Kernel Fuzzy C-Means Algorithm
V. Mageshwari, I. Laurence Aroquiaraj, T. Dharani

Abstract: Images are imitations of factual world substances. Processing it to get better visualization is called as image processing. With the increasing availability and decreasing cost of satellite imagery, the Remote sensing image enhancement, segmentation and classification has become the most important research issue in field of Remote sensing. In this proposed work, Land sat 7 Remote Sensing images are considered. Initially the enhancement of satellite image is done using image enhancement techniques. Then the segmentation of satellite images has been done using Expectation Maximization (EM), Kernel-Means (K-Means), Kernel Fuzzy C-Means (KFCM) and Modified Kernel Fuzzy C-Means (MKFCM) algorithms. Results are obtained for different Land sat 7 Remote Sensing images. Finally quality measures such as mean square error, average difference, normalized cross correlation and error measurements like Peak signal to noise ratio, Normalized absolute error are calculated.