JOURNAL OF SOFT COMPUTING AND INFORMATION TECHNOLOGY (JSCIT)
Year:2016 | Volume:5 | Issue:3 |Papers Count:7

The increasing use of Multi-Tier Application (MTA) in virtualized environments necessitates performance evaluation of such systems to achieve scalable and flexible services. However, providing appropriate performance for Virtualized Multi-Tier Applications (VMTA) that have complex architecture is much more difficult than traditional application architecture. In this paper, we propose a placement strategy to settle virtual machines (VMs) of VMTAs in virtualized datacenters. First, VMs hosting tiers are ranked based on Cobb-Douglas production function. Then, VMTAs are prioritized based on resources utilization and performance metrics of hosts. Finally, VMTAs are placed regarding resource demands of tiers and their functional dependency. Results reveal that the proposed solution excels in terms of load balancing and energy consumption, while reduces Service Level Agreement (SLA) violation and VMs interference in the datacenter.

Today using the internet has spread wildly, and increasing number of web pages leads to importance of using search engines, therefore some people try to misguide search engines to have more customers and benefit. They increase the rank of their pages by some illegal ways. search engines to. Identify of this kind of web pages can improve search engines and attract confidence to user. According to importance of finding spam pages, the research is presented a new linke-based way to detect spam pages in Persian web graph. This way, first link farms detectes. Finally, the negative scores of spam pages propagate in whole of web graph. This way was implemented on data of Parsijoo search engine and the result of data analyses indicates 21.2% improvement in p@n factor.

Web users Clustering involves finding a structure and pattern in a series of web events and leads to personalized pages, recommender systems and direct marketing in e-commerce. It's always important in clustering areas to extract Clusters with the shortest intra-cluster distance and greatest inter-cluster distance. This paper presents an algorithm for clustering web users using a combination of bacteria algorithms and cultural algorithms. In this regard the problem space is modeled using Bacterial Foraging Optimization Algorithm. Then a cultural space is created for the problem that maintains proper norms of spots. The created cultural Space is effective to perform better bacteria evolutionary operations such as gene division and gene integration and prevents accidental operations. Two real data sets, EPA and NASA, are used to tests that. The results indicate better performance of this method compared with other algorithms.

Task scheduling has been so far of important challenges in high-performance computers e.g. parallel and distributed systems. Using such architectures during compiling, each application program is divided to some tasks. Because of data-flow among the tasks, they may be dependent to one another; hence, there will be precedence constraints and communication delays among them so that each application with its corresponding tasks can be modeled using a Directed Acyclic Graph (DAG) named task graph. In static task-graph scheduling in homogeneous multiprocessor environments, tasks in the given task graph should be mapped to a predefined number of identical processing elements regarding the precedence constraints and communication delays so that the program’s completion time (finish time) is minimized, and this is an NP-hard problem form the time-complexity perspective. Actually, the achieved results are dominated by two different-in-nature factors: 1) which topological order of tasks should be considered? (sequencing subproblem), and 2) how should the extracted order be distributed over the processors? (assigning subproblem). In this paper, an efficient hybrid approach is proposed in which the Ant Colony Optimization (ACO) determines the order of tasks, and a Cellular Learning Automata (CLA) machine tackles with the assigning subproblem, and maps the task order derived by ACO to the existing processors.125 randomly-generated task graphs with different shape parameters such as size, Communication-to-Computation Ratio (CCR), and parallelism are used for the comparison study, and the results shows that the proposed approach is more successful than the traditional counterparts from the performance point of view, and eventually outperforms them.

Since there is no fixed infrastructure or centralized management in Wireless Sensor Networks (WSNs), a Connected Dominating Set (CDS) has been proposed as a virtual backbone. In this paper, degree-constrained minimum-weight connected dominating set (DC-MWCDS) problem is used for modeling energy-efficient backbone formation in wireless sensor networks in UDG. DC-MWCDS aims at forming degree-constrained backbone and minimizing the weight of the CDS made for network graph simultaneously. The aim of this paper, is proposing an energy efficient connected dominating set (CDS) scheme in wireless sensor networks, which prolongs the network lifetime. In proposed algorithms, we use an optimal weight based on the minimum residual energy and maximum effective degree of nodes for backbone formation to prolong the network lifetime. The optimal weight coefficients are determined using particle swarm optimization (PSO) algorithm. Then, when selecting nodes for dominating set (DS) formation, these coefficients will be used. If the degree of a node is more than coefficient of degree constraint and energy of a node is less than coefficient of energy constraint, the node won’t be selected for DS formation. The message and time complexity of the proposed algorithm is O (n). Simulation results show that proposed algorithms outperforms the other methods in terms of network lifetime.

Due to inherent limitations and number of distributed sensor nodes in wireless sensor networks (WSNs), the number of generated packets will be increased and congestion will be produced in link level or node level. Hence, the number of lost packets will be increased and the network resources such as energy will be consumed in inappropriate manner. So, the congestion is one of the most important challeng in WSNs. In this paper, a method is provided for controlling congestion. For congestion detection, the queue lenghth will be used and the occupied level of node’s buffer is indication of congestion level. In the proposed scheme, when the congestion occures, the congested node informs its upstream neighbour nodes. In order to controlling congestion, the upstream nodes will be reduced the transmitted pacekt rate and selects another path for data transmission. In selecting secondary path for data transmission node considers parameters such as amount of available energy and the empty size of the buffer. Simulation results show that the proposed scheme improve the consumtion energy and packet deliver rate in compraition with PCCP [9].

In this paper, a downlink system with cellular characteristics and two-way frequency division is considered which each user sends the information of status of information chanel to the home station. Since the collaboration between all the stations practically is not possible, a number of stations are classified into the clusters. Users should estimate their channel by sending pilots to the center station, and sending feedback of channel information. On the other hand, the station refuses to send the information, until it recives all information of users. This action leads to delation in transmission of information in downlink. In this research, four methods for reducing the information feedback of channel status in downlink system are proposed. The main idea of the proposed methods including the division the users into two groups based on the power of user channels and user distances from the transmitter station. The key issue in CoMP is that channel status information should be collected in a small fraction of the channel's binding time, so that it has enough time to send information. Using simulation, it is observed that the effect of the proposed methods to reduce the feedback of channel status information on system performance is very low compared to other methods.