International Journal of Computer Information Systems, Vol. 3, No.

3, 2011

Locating Objects efficiently in a Distributed System using Particle Swarm Optimization

S.Raghavendran
2

Assitant professor, Computer Science Engineering, Vel Tech DR.RR & DR.SR Technical University
Chennai, India [email protected]

S.Arun

Assitant professor, Electrical Engineering, Vel Tech High Engneering College, Chennai,India
[email protected]

AbstractAs distributed systems can solve computational problems it can coordinate the use of shared resources and also provide communication services to the users. The server automatically re-configures the routing information when a new node is attached to the network. This project aims to locate the objects in the distributed system. The name of the object is broadcasted to the network by means of Ant Colony Optimization (ACO) algorithm. Once the object is identified Particle Swarm Optimization (PSO) algorithm blocks broadcasting messages to further levels. PSO algorithm also update the object name and i/p address of the peer in the routing table. For new search PSO algorithm checks the routing table for the Object, if present directly forwarded to the peer there by reducing the time delay and traffic over the network Keywords- Object-based distributed systems, Optimization, Particle Swarm Optimization. Ant Colony

The Particle swarm optimization (PSO) algorithm also resolves the i/p address of the peer where the object is identified and sends a response to the server there by updating the routing table with object name and its corresponding node. For a new search of an object, PSO algorithm initially checks the routing table for availability. If it is available, directly forward to the peer and locate the objects efficiently there by reducing the time delay and traffic over the network. II. SYSTEM DESIGN

I.

INTRODUCTION

Distributed system consists of multiple autonomous computers that communicate through a computers network. Object Oriented Programming (OOP) model has become the most popular approach which involves large application and huge amount of data. In a distributed system, object can be distributed to all the peers. Identifying peer node in a distributed system and to transfer the message between the nodes for parallel computation and interprocess communication. Routing information of the network is maintained in a separate node. Any node attached to the network is reconfigured in the routing table present in the server. For accessing the object concern every object is referenced through its symbolic name. The name of the object should be independent. The object name is broadcasted through the network by means of ACO algorithm. The neighbour connected to the peer can also receive the message along with the object in the header. All the peers are simultaneously allowed to search for the object (file name) by query receiver. If present, further broadcasting of message will be deterred. If the object is not identified the successive neighbour will receive the broadcast message by means of query forwarder present in the antecedent. Once the object is identified PSO algorithm blocks broadcasting the messages to further levels.

III.

NETWORK CONSTRUCTION

A network construction allows sharing of resources and information among devices connected to the network. Clientserver networking is a distributed application architecture that partitions task or work loads between service provider (server) and service requester (client). Offer clients and server operates over a computer network on separate hardware. Client-server architecture enables the role and responsibilities of a computing system to be distributed among several independent computers that are known to each other only through a network. In the server side construction, a server can maintain the routing table to store the information of the peer and name

Special Issue I

Page 9 of 55

ISSN 2229 5208

of the object. In the client side construction of network, it contains query forwarder, query receiver and file content. IV. ACO AND ACO-PSO

International Journal of Computer Information Systems, Vol. 3, No. 3, 2011 The transition rules, the local and global pheromone updating rules are consecutively defined in equation (1-3) Pm (i, j) = if j allowedk Otherwise (i ,j)(1-) (i ,j)+( i , j) (i ,j)(1-) (i ,j)+( i , j) (2) (3) (1)

The ant colony optimization (ACO) [2-3] is inspired by the behaviors of real ants. An ant leaves some pheromone at the amount of on ground when it travels and marks the path by a trail of this substance. The pheromone would evaporate at a certain rate as time goes on. The next ant will smell the pheromone remained on different paths and chooses a one with a probability proportional to the amount of pheromone. The ant that follows the path will leave its own pheromone. This phenomena is considered as a positive feedback process which could be treated as knowledge sharing through collaborative efforts. Based on previous knowledge, ants progressively construct their tours. The best tour reported by this algorithm is a path in the graph that reflects the attempt to optimize TSP objective. The components of ACO which are used to build up the system are , , and Q0 . and respectively control the relative importance of pheromone trail and distance between the cities in TSP. refers to the rate of pheromone evaporation. Q0 is the major component of ACO. Such a parameter regulates chances of exploitation or exploration as aforementioned. The ACO algorithm and the associated parameters are mentioned. // Initialization phase 1. Represent the underlying problem by a connected weighted graph. 2. Set the parameters: Initialized amount of pheromone for each edge; the value of relevant parameters i.e., , , ; the global best fitness to BIG_NUMBER (i.e., 109) // End of initialization phase // Evolutionary phase 3. Repeat 3.1 For each ant do 3.1.1 Randomly select a starting node 3.1.2 Repeat 3.1.2.1 Move to the next node according to the node transition rule, given , and Q0 . 3.1.2.2 Update the pheromone intensity using local pheromone updating rule, given . Until a tour is completed. End For 3.2 Update best fitness. 3.3 For each edge do Update the pheromone intensity using global pheromone updating rule, given . End for Until MAX_ACO run is attained // End of evolutionary phase ACO algorithm

Where (i, j) is the quantity 1/dij, allowedk is {N tabuk}, ( i , j) is the quantity per unit of length of trail substance lain on edge (i,j), is a coefficient such that (1- ) represents the evaporation of trail in global pheromone updating rules. is a coefficient such that (1- ) represents the evaporation of trail in local pheromone updating rules. According to ACO algorithm, commands in 3.1.2 refer to the ACO local search. The search is computed by using equation (1) and (2). In 3.1.2.1, Q0 is one of the parameter involved with such a search process. If Q0 value is high, the local search would turn towards exploitative type. Such a type of the search concerns with the selection of the node associated with the maximum value of Pij regarding equation (1). On the contrary, if the value of Q0 is low, the explorative type of the search is taken into account. With respect to the latter search type, the selection would be made on any random node associated with Pij that is greater than the mean of the choices. The modifications of ACO [2] proposed here mainly concentrate on the parameter Q0 which influences ACO local search as aforementioned. ACO-PSO is one of modifications of ACO performed in this paper. The particle swarm optimization (PSO) could be one of the possible choices to adjust Q0 in ACO. In PSO, a population of particles adapts by returning stochastically toward previously successful regions in the search space. Each particle i in a population has a position, which defines a potential solution. It also has a velocity, . Each particle remembers its . or neighborhood best previous best position found, position, . The particle would be able to access to the global . At each iteration, a particles velocity is best position,

updated using its current velocity, the known previous best position and the global best position. The new particle position is then calculated by simply adding the velocity to the current position. In this paper, an inertia weight w is employed to control the impact of the previous velocity on the current velocity. The w provides a mechanism for forcing the swarm to perform a global search, maintaining diversity in the population, or a local search, reducing diversity in the population. Typically w is reduced linearly during the run in order to facilitate a global search early in the run and a local search towards the end of a run. Thus: w (k) = w0 - (w0 - w1) (4)

Special Issue I

Page 10 of 55

ISSN 2229 5208

Where w0 is the initial inertia weight and w1 is the inertia weight in the last iteration, with 0 < w1 < w0, k is the current number of iterations. K is the maximum number of iterations. The particles are manipulated according to the following set of equations, equation (5) and (6): (t) = w* ( t - 1) + c1 * 1 * ( + c2 * 2 * ( (5) (6) t represents the current iteration. c1 and c2 are two positive constants. 1 and 2 are two uniform random numbers in the range [0,1]. For each iteration, each particles velocity and position are updated using equation (5) and (6) respectively. The particles position is then adjusted to reflect the new velocity. The maximum and minimum value of is given as boundaries for preventing too much divergence of the solutions. ACO-PSO could be one of the modifications of ACO by using PSO. The method of ACO-PSO is described in ACOPSO algorithm. In this algorithm, it is not necessary that the position and velocity are vectors. They can be represented by scalars since they relate to Q0 which is a real-value element. // Initialization phase 1. For each particle i 1.1 Randomly initialize xi , vi for particle i Each xi corresponds to a particular value of Q0 End For // End of initialization phase // Evolutionary phase 2. Do 2.1 For each particle i 2.1.1 Call ACO algorithm to produce the fitness of the particle 2.1.2 If current_fitness_value is better than previous_best_fitness_value ( pi ) Then current_fitness_value of particle i becomes pi End If End For 2.2 Call find_global_best_fitness 2.3 Check if it gets trapped in a local minimum or not // The local minimum state is indicated in the next section 2.3.1. If it is; Then generate the new population of Chromosome with the value in a range of [0, 0.5] And go to equation (1) 2.3.2 If it is not; Then go to equation (4) 2.4 For each particle i 2.4.1 Call calculate_ vi based on equation (2) 2.4.1 Call calculate_ xi based on equation (3) End For Until MAX_PSO run is attained // End of evolutionary phase ACO-PSO algorithm

International Journal of Computer Information Systems, Vol. 3, No. 3, 2011 V. ROUTING TABLE INFORMATION A search protocol and a routing table updating protocol are further proposed in order to expedite the search process through self organizing P2P network. A node first checks whether the searching object is present in the routing table. If its available directly forwarded to the node and identify the object. The routing table contains peer i/p address and name of the object. A search protocol and a routing table updating protocol are further proposed in order to expedite the search process through self organizing network. A node first checks whether the searching object is present in the routing table. If it is available, directly forwarded to the node and identify the object. The routing table contains i/p address of the peer and name of the object. VI. 1. 2. 3. 4. 5. 6. 7. ROUTE DISCOVERY WORKING PROCEDURE:

Setting the neighbouring nodes for each system. Request for data transmission is made. Broadcast the request message to all the neighbouring nodes until the destination is reached. Each of the intermediate nodes adds its address to the request message. The destination node replies with the reply message to the source through the intermediate nodes. The source now stores the path in its route cache. This table contains the node name and it is i/p address connected in the server.

This table contains the information of the object searched in the distributed system.

This table maintains the information of the neighbouring node.

VII.

QUERY FORWARDER

The name of the object is broadcasted to the network by means of Ant Colony Optimization (ACO) algorithm. The neighbour connected to the peer can also receive the message along with the object in the message header. If the object is not identified the request is broadcasted to further levels in the network. Once the object is identified, PSO algorithm blocks

Special Issue I

Page 11 of 55

ISSN 2229 5208

International Journal of Computer Information Systems, Vol. 3, No. 3, 2011 broadcasting the messages and resolves the i/p address of the peer where the objects is identified and send a response to the server for updating the routing table. VIII.
QUERY RECEIVER

Query receiver present in every node is responsibility for search for the object (file name) and if present further broadcasting of message will be defended. If the object is not identified, the successive neighbour will receive the broadcast message. IX. EXPERIMENTAL ANALYSIS

Figure 4 Node completed connection

The name of the client is entered into the system. The name is checked with routing table in the server for availability.

IX. I. Server Side A network construction allows sharing of resources and information among devices connected to the network. Clientserver networking is a distributed application architecture that partitions task or workloads between service provider (server) and service requester (client).All the nodes are to be connected to the server. The following screen shot specifies the number of nodes to be connected to the server.
Figure 5 Node login

Figure 2: Specifying the number of nodes

Name each corresponding client nodes connecting to the server. Naming node establish the connection with the server. Client nodes are registered to the server. For every client, entry is made in the routing table.

The name of the object is broadcasted to the network by means of Ant Colony Optimization (ACO) algorithm. The neighbors connected to the peer can also receive the message along with the object in the message header. Each and every search of an object is stored in the routing table. Once the object is identified, PSO algorithm blocks broadcasting the messages and resolves the i/p address of the node where the object is present and sends a response to the server for updating the routing table. File sharing is providing the access to the nodes in the network objects. Before searching objects in the distributed network, all the objects are shared to the distributed system. File sharing is the practice of distributing or providing access to digitally stored information, such as computer programs, multi-media (audio, video), documents, or electronic books. It may be implemented through a variety of storage, transmission, and distribution models. In a file sharing environment, a large number of users can access a program as though it were on their local machines, when actually the program resides on a single file server. This is a great benefit to small workstations, where disk space is at a premium. A user can have access to a much larger program repertoire than could fit on a private disk.

Figure 3 Node registrations to the server

Establishing the connection between the nodes is used to identify the neighbouring node. Connection between the nodes can be implemented

Figure 6 file sharing to the nodes

Special Issue I

Page 12 of 55

ISSN 2229 5208

International Journal of Computer Information Systems, Vol. 3, No. 3, 2011 After sharing the files, display the node name on which the file is shared. The name of the object is broadcasted to the network by means of Ant Colony Optimization (ACO) algorithm for the first search of an object. Once the object is identified, Particle Swarm Optimization (PSO) algorithm blocks broadcasting messages to further levels. PSO algorithm also update the object name and i/p address of the peer in the routing table. For new search PSO algorithm checks the routing table for the Object, if present directly forwarded to the peer there by reducing the time delay and traffic over the network. Searching the object in the system, initially checks the routing table where the object is identified and subsequently PSO algorithm is activated. Else the ACO algorithm will search the entire network for the object. Searching the object from all nodes connected to the server and displaying the node name where the object resides.

REFERENCE [1]
Jiabin Li and Chun Che Fung: Locating Object Efficiently in a Distributed Computing System using Ant Colony Optimisation (IEEE DEST 2008) [2] Sarayut Nonsiri and Siriporn Supratid: Modifying Ant Colony Optimization presented in 2008 IEEE Conference on Soft Computing in Industrial Applications (SMCia/08), June 25-27, 2008, Muroran, JAPAN. [3] M. Dorigo, M. Birattari, and T. Stutzle, "Ant colony optimization: artificial ants as a computational intelligence technique," Universite Libre de Bruxellest, Belgium 2006. [4] S. Ratnasamy, P. Francis, M. Handley, R. Karp, and S.Shenker, "A Scalable Content-Addressable Network," presented at ACM SIGCOMM, 2001. [5] A. Kumar, J. Xu, and E. W. Zegura, "Efficient and Scalable Query Routing for Unstructured Peer-to-Peer Networks," presented at INFOCOM 2005. 24th Annual Joint Conference of the IEEE Computer and Communications Societies, 2005. [6] K. Gummadi, R. Gummadi, S. Gribble, and S. Ratnasamy, "The impact of DHT routing geometry on resilience and proximity," presented at ACM SIGCOMM'03, 2003. [7] G. S. Manku, M. Bawa, and P. Raghavan, "Symphony: Distributed hashing in a small world," presented at 4th USENIX Symposium on Internet Technologies and Systems, 2003. [8] M. Castro, P. Drschel, Y. C. Hu, and A. I. T. Rowstron, "Topology-aware routing in structured peer-to-peer overlay networks," presented at International Workshop on Future Directions in Distributed Computing, 2003. [9] H. Zhang, A. Goel, and R. Govindan, "Incrementally improving lookup latency in distributed hash table systems," presented at ACM SIGMETRICS 2003, 2003. [10] C. C. Wang and K. Harfoush, "RASTER: a light-weight routing protocol to discover shortest overlay routes in randomized DHT systems," presented at 12th International Conference on Parallel and Distributed Systems, 2006 AUTHORS PROFILE

Figure 7 Node completed connection Displaying the node name and object present in the distributed system.

Mr. S.Raghavendran, He is working as Assistant Professor in Vel Tech Dr.RR Dr.SR Technical University,Chennai and Finished M.E., (CSE), Sri Sivasubramaniam College of Engineering, Affiliated to Anna university in 2010 and Finished his B.E (CSE) in Saveetha Engineering college, Affiliated to Anna university in 2008 and He presented papers in several national and international conferences. Mr. S.Arun was born on 18th october 1979 Tiruvanamalai, India. He is working as Assistant Professor in Vel Tech High Engneering College Chennai and pursuing Ph.D. in the School of Electronics and Electrical Engineering, Singhania University, Rajasthan, India. He obtained M.E. degree in Communication systems 2004, Anna University and B.E. degree in Electronics and Communications Engineering from University of Madras, Chennai, India in the year 2001. . He has also published papers on image processing, wireless communication in National and International Conferences. His current research interests are Robotics, wireless Communications.

Figure 4.14 locating the required object Saving the object present in the distributed system in the appropriate location of the system Searching the object (from node a) and the searching process is implemented in the distributed system and display, where the object is present (present in node b). X. CONCLUSION

This paper describes a methodology to identify object in a distributed network of varying complexity through the use of particle swarm optimization algorithm and Ant colony optimization algorithm (ACO) algorithm. This reduces the time delay in finding an object with less traffic overhead

Special Issue I

Page 13 of 55

ISSN 2229 5208