Welcome to Scribd, the world's digital library. Read, publish, and share books and documents. See more ➡
Download
Standard view
Full view
of .
Add note
Save to My Library
Sync to mobile
Look up keyword
Like this
1Activity
×
0 of .
Results for:
No results containing your search query
P. 1
Implementation of a new Fuzzy Based Load Balancing Algorithm for Hypercubes

Implementation of a new Fuzzy Based Load Balancing Algorithm for Hypercubes

Ratings: (0)|Views: 638|Likes:
Published by ijcsis
Distributed computing systems are becoming increasingly available because of the rapid decrease in hardware cost and the advances in computer networking technologies.. It is frequently observed that in a computing environment with a number of hosts connected by a network, the hosts are often loaded differently. In typical distributed system task arrive at the different nodes in random fashion. This causes a situation of non-uniform load across the different nodes. Load imbalance is observed by existence of nodes that are highly loaded while the others are lightly loaded or even idle. Such situation is harmful to the system performance in terms of response time and resource utilization. In the work presented in this paper we have tried to analyze the effect of using fuzzy logic to deal with the problem of load balancing in hypercube model.
Distributed computing systems are becoming increasingly available because of the rapid decrease in hardware cost and the advances in computer networking technologies.. It is frequently observed that in a computing environment with a number of hosts connected by a network, the hosts are often loaded differently. In typical distributed system task arrive at the different nodes in random fashion. This causes a situation of non-uniform load across the different nodes. Load imbalance is observed by existence of nodes that are highly loaded while the others are lightly loaded or even idle. Such situation is harmful to the system performance in terms of response time and resource utilization. In the work presented in this paper we have tried to analyze the effect of using fuzzy logic to deal with the problem of load balancing in hypercube model.

More info:

Published by: ijcsis on Jan 20, 2011
Copyright:Attribution Non-commercial

Availability:

Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See More
See less

01/20/2011

pdf

text

original

 
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No.9, 2010
 Implementation of a new Fuzzy Based Load  Balancing Algorithm for Hypercubes
Sameena Naaz Afshar Alam Ranjit BiswasDepartment of Computer Science, Jamia Hamdard Institute of Technology and ManagementHamdard University New Delhi, India Gurgaon, Haryanasnaaz@jamiahamdard.ac.in aalam@jamiahamdard.ac.inranjitbiswas@yahoo.com
 Abstract 
 — 
 Distributed computing systems are becoming increasingly available because of the rapid decrease inhardware cost and the advances in computer networkintechnologies.. It is frequently observed that in a computing environment with a number of hosts connected by a network,the hosts are often loaded differently. In typical distributed  system task arrive at the different nodes in random fashion.This causes a situation of non-uniform load across thedifferent nodes. Load imbalance is observed by existence of nodes that are highly loaded while the others are lightlyloaded or even idle. Such situation is harmful to the system performance in terms of response time and resourceutilization. In the work presented in this paper we have tried to analyze the effect of using fuzzy logic to deal with the problem of load balancing in hypercube model 
 Keywords Load Balancing, Fuzzy Logic, Hypercubes, ResponseTime
I. I
 NTRODUCTION
 With the rapid decrease in the cost of hardware and thesimultaneous increase in the computer networking technologiesthe increase of distributed computer systems has increased alot. The obvious advantage of using these systems isinformation and resource sharing. At the same time we knowthat it allows parallel execution of a job on multiple processors.When the jobs are being executed in parallel on differentsystems a decision has to be made on to which system a newlyarrived job has to be send. In a typical distributed system the jobs arrive in random fashion on different nodes. This causes asituation where some of the nodes are heavily loaded whereasothers are lightly loaded. Load balancing is the techniquewhich helps in even distribution of the jobs among theavailable nodes so that the throughput and the response timescan be increasedDifferent load balancing algorithms have differentcomplexity which depends upon the amount of communicationneeded to approximate the least loaded node. These algorithmscan be static or dynamic in nature. Static algorithms collect noinformation and make probabilistic balancing decisions, whiledynamic algorithms collect varying amounts of stateinformation to make their decisions. Previous research on staticand dynamic load balancing can be found in [l]-[5], [6, 7],respectively. It has been established from the previous studiesthat dynamic algorithms give better performance improvementas compared to static algorithms.When we are talking about large distributed systems thereis huge amount of global state uncertainty present in it. Fuzzylogic based distributed load balancing algorithms reflect theeffect of uncertainty in decision making process. This approachhas been discussed in [8]. The fuzzy logic approach foDistributed Object Computing Network has been studied in [9,10]. Parallel and distributed computing environment isinherently best choice for solving/running distributed and parallel program applications. In such type of applications, alarge process/task is divided and then distributed amongmultiple hosts for parallel computation. In [10] it has been pointed out that in a system of multiple hosts the probability of one of the hosts being idle while other host having multiple jobs queued up can be very high. In [11] the performance of anew Fuzzy Load balancing algorithm is compared with theexisting algorithms.In a distributed environment the processors are categorizedaccording to workload in their CPU queues as heavily loaded(more tasks are waiting to be executed), lightly loaded (lesstasks are waiting to be executed in CPU queue) and idle processors/hosts (having no pending work for execution). HereCPU queue length is used as an indicator of workload at a particular processor. The algorithms used for load balancingmay require no information, or only information aboutindividual jobs (static algorithm) or may make decisions basedon the current load situation (dynamic algorithm).In general, load balancing algorithm can be analyzed in aframework with four dimensions: selection policy, transfer  policy, information policy, and location policy. Specifically,information and location policies have the most importantroles.
Transfer policy
: First of all the state of the differentmachines is determined by calculating it’s workload.A transfer policy determines whether a machine is ina suitable state to participate in a task transfer, either as a sender or a receiver. For example, a heavilyloaded machine could try to start process migrationwhen its load index exceeds a certain threshold.
Selection policy
: This policy determines which task should be transferred. Once the transfer policydecides that a machine is in a heavily-loaded state,the selection policy selects a task for transferring.Selection policies can be categorized into two policies: preemptive and non-preemptive. A preemptive policy selects a partially executed task.
270 http://sites.google.com/site/ijcsis/ISSN 1947-5500
 
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No.9, 2010
As such, a preemptive policy should also transfer thetask state which can be very large or complex. Thus,transferring operation is expensive. A non- preemptive policy selects only tasks that have not begun execution and, hence, it does not requiretransferring the state of task.
Location policy
: The objective of this policy is tofind a suitable transfer partner for a machine, oncethe transfer policy has decided that the machine is aheavily-loaded state or lightly-loaded one. Commonlocation policies include: random selection, dynamicselection, and state polling.
Information policy
: This policy determines when theinformation about the state of other machines should be collected, from where it has to be collected, andwhat information is to be collected.II.D
ISTRIBUTED
S
YSTEM
M
ODEL
A simple model of a distributed system is presented here.This model consists of a decentralized decision makingapproach with cooperation from all the nodes. So the performance can be improved here purely by intelligentdecision making and proper coordination. The various nodesof the system here are the resources and each of theseresources can be in different states. A vector is used to givethe state of a node which describes many characteristics of thenode. The elements of this state vector are measures whichimply a cost or penalty for using the resource.The set of states of all the resources in the distributedsystem is known as the global system state. In distributed load balancing also the decisions are not always necessarily madeusing the complete global state information. In fact for eachnode under consideration only a subset of neighboring nodesmay be needed to take a decision. Another important aspect isthat a node can change state faster than the time taken totransmit state information from one state to another. Thereforethere is always some amount of uncertainty in the stateinformation used for making a decision. Hence it is necessarythat the decision making process deals with theseuncertainties. Fuzzy logic is one of the methods of dealingwith this uncertain information and has been used in the work  presented in this paper.III.D
ESCRIPTION
 
OF
A
LGORITHM
The algorithm implemented in this work is basically dividedinto two parts:A.The system ModelB. The Scheduler a)Threshold Estimation b)Decision Making
 A.The System Model 
A distributed system is assumed to be c collection of autonomous nodes connected by a communication network.Message passing is the only form of communication amongnodes. The system model with d dimensions has 2^d nodes. If we number the nodes from 0 thru (2^d)-1 and look at thenumbers as d-digit binary numbers, then each node will beconnected to d other nodes which differ only in one bit. Thatis, suppose d=4 for example, then 0010 (node 2) will beconnected to 1010 (node 10), 0110 (node 6), 0000 (0) and0011 (node 3).The system model is a compound module typewhich consists of a sender (source) and receiver (sink)module.
 B. The Scheduler 
Scheduler has two functions, threshold estimation and decisionmaking. When a scheduler is invoked, it estimates twonumerical thresholds from the current states of uncertaintysources based on a fuzzy control base, and making schedulingand state update decision using fuzzy consistency model.We have done the implementation of scheduler on MATLAB.In fuzzy logic toolbox we have take two input parameters. Thefirst input parameter is ‘load’ and the second one is ‘Number of heavy Load Node’ and one output i.e. ‘status of load balance node’. We measure load and Number of heavy loadnode on a 0 to 10 scale and status of load balancing node on 0to 20 scale.We need to define fuzzy sets for the input parameters, ‘load’,and ‘number of heavy load node’ levels, and the output, ‘statusof load balancing node’. For this we define five membershipfunctions for first input parameter i.e. ’load’ and twomembership functions for second input parameter i.e. ‘number of heavy load node’ and two membership functions for output parameter ‘status of load balance node.’
Threshold Estimation
The Threshold Estimation determines the limiting value for each membership function. Beyond this limiting value themembership function will change.
First Input parameter: Load (0-10)
Member Function 1: Very lightly (0-2)Member Function 2: lightly (1-5)Member Function 3: moderate (4-6)Member Function 4: heavy (5-9)Member Function 5: very heavy (8 -10)
Second Input Paramete
r:
No. of heavy load node (0-5)
Member Function 1: more (0-2.5)Member Function 2: less (2.5 5)
Output Parameter:
 
Status of load balance node (0-10)
Member Function 1: receiver (0-5)Member Function 2: sender (6-10)In our work here we have taken the Gaussian distributionfunction for all the different linguistic variables for theinput “load”. This is shown in figure 1 below.
271 http://sites.google.com/site/ijcsis/ISSN 1947-5500
 
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No.9, 2010
Figure 1: Input variable load of the node undeconsideration and it’ membership functionThe membership function used for the number of heavyload nodes is shown in figure2.Figure 2: Input variable No. of Heavy Load Node and it’smembership function.Figure 3: Membership function for the output variableStatus of Load Balance NodeThe membership function for the output variable status of load balance node is shown in figure 3. From this figurewe can see that there are two linguistic variables sender and receiver here and the load on a node determines it’svalue based upon the membership function.
 Decision Making 
The Fuzzy rules that have been used in this work aregiven below:Rule [1]. If (load is very light) then (node is receiver)Rule [2]. If (load is very heavy) then (node is sender)Rule [3]. If (load is heavy) and (no. of heavy load nodes isless) then (node is sender)Rule [4]. If (load is heavy) and (no. of heavy load nodes ismore) then (node is receiver)Rule[5]. If (load is light) and (no. of heavy load nodes ismore) then (node is receiver)Rule[6]. If (load is light) and (no. of heavy load nodes isless) then (node is sender)Rule [7]. If (load is moderate) and (no. of heavy loadnodes is more) then (node is receiver)Rule [8]. If (load is moderate) and (no. of heavy loadnodes is less) then (node is sender)
272 http://sites.google.com/site/ijcsis/ISSN 1947-5500

You're Reading a Free Preview

Download
/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->