Load Balancing (Synopsis)

Published on May 2016 | Categories: Types, School Work | Downloads: 14 | Comments: 0 | Views: 95
of 8
Download PDF   Embed   Report

ieee projects titles,ieee projects list,ieee projects download,project titles in java,ieee projects on java,ieee projects for cse,ieee project,project ideas in java,ieee projects for ece,mba project list,mba project topics,mca project titles,project topics for computer science,java projects for engineering students,ieee projects for eee,mca project topics,topics for projects in java,computer project topics,java project topics,computer science project topics,ieee project titles,ideas for projects in java,ieee project titles for cse,ieee project topics for ece,ieee projects titles for cse,projects ideas in java,ieee projects titles 2010,ieee projects title,project topics for information technology,ieee project topics,ieee projects in java,ieee projects list 2010,mechanical project topics,final year project titles,mechanical projects ideas,ieee java projects 2010,mechanical project ideas,project topics in java,java projects ideas,engineering project topics,mba project titles,eee projects,projects for ece,ieee projects,mba projects list,java projects download,mba project,project for computer science,computer engineering projects,list of mba projects,topics for project,ieee project titles 2011,ece projects list,mba projects topics,mba projects download,ieee project titles for ece,project title in java,topics for java project,final year projects for information technology,mca projects in java,java mini project topics,,java projects topics,project for ece,projects titles in java,ece projects titles,java project topics list,topics of projects,java ieee projects,projects for computer science,download mba projectsieee it projects,mba project topics in finance,mba marketing project topics,list java projects,mca final year projects titles,mini project titles in java,it projects in java,mba project topics for marketing,mba projects in finance,new project topics in java,project ideas for engineering students,final year projects titles,project list in java,marketing topics for mba project,students projects,eee projects titles,mba projects,mba project topics in marketing,topics for project in java,projects in ieee,list of java projects,mba project title in marketing,mini project topics for mechanical,marketing project topics for mba,project topics for eee,topics for mini ,projects in java,java projects for mca,project for final year,mca project ideas,java project titles,ieee projects titles,ieee projects list,ieee projects download,project titles in java,ieee projects on java,ieee projects for cse,ieee project,project ideas in java,ieee projects for ece,mba project list,mba project topics,mca project titles,project topics for computer science,java projects for engineering students,ieee projects for eee,mca project topics,topics for projects in java,computer project topics,java project topics,computer science project topics,ieee project titles,ideas for projects in java,ieee project titles for cse,ieee project topics for ece,ieee projects titles for cse,projects ideas in java,ieee projects titles 2010,ieee projects title,project topics for information technology,ieee project topics,ieee projects in java,ieee projects list 2010,mechanical project topics,final year project titles,mechanical projects ideas,ieee java projects 2010,mechanical project ideas,project topics in java,java projects ideas,engineering project topics,mba project titles,eee projects,projects for ece,ieee projects,mba projects list,java projects download,mba project,project for computer science,computer engineering projects,list of mba projects,topics for project,ieee project titles 2011,ece projects list,mba projects topics,mba projects download,ieee project titles for ece,project title in java,topics for java project,final year projects for information technology,mca projects in java,java mini project topics,,java projects topics,project for ece,projects titles in java,ece projects titles,java project topics list,topics of projects,java ieee projects,projects for computer

Comments

Content


Load Balancing

Abstract
The approach considers the heterogeneity in the processing rates of
the nodes as well as the randomness in the delays imposed by the
communication medium. The optimal one-shot load balancing policy is
developed and subsequently extended to develop an autonomous and
distributed load-balancing policy that can dynamically reallocate
incoming external loads at each node. This adaptive and dynamic load
balancing policy is implemented and evaluated in a two-node
distributed system. The performance of the proposed dynamic load-
balancing policy is compared to that of static policies as well as
existing dynamic load-balancing policies by considering the average
completion time per task and the system processing rate in the
presence of random arrivals of the external loads.
.
Introduction
About the Project
A sender-initiated distributed DL policy where each node
autonomously executes L at every external load arrival at that node.
The DL policy utili!es the optimal one-shot L strategy each time an
L episode is conducted" and it does not require synchroni!ation
among nodes. #very time an external load arrives at a node" only the
receiver node executes a locally optimal one-shot-L action" which
aims to minimi!e the average overall completion time.
This requires the generali!ation of the regeneration-theory-based
queuing model for the centrali!ed one-shot L. $urthermore" every L
action utili!es current system information that is updated during
runtime. Therefore" the DL policy adapts to the dynamic environment
of the distributed system.
ene%ts of Dynamic load balancing &
 The scheduling decisions are made at runtime
 L Threshold value is dynamic
 'erver allotting process is dynamically predetermined
 Download process is very fast
System Analysis
Existing System
we have shown that" for distributed systems with realistic
random communication delays" limiting the number of balancing
instants and optimi!ing the performance over the choice of the
balancing times as well as the L gain at each balancing instant
can result in signi%cant improvement in computing e(ciency.
This motivated us to look into the so-called one-shot L strategy.
)n particular" once nodes are initially assigned a certain number
of tasks" all nodes would together execute L only at one
prescribed instant . *onte +arlo studies and real-time
experiments conducted over ,LA- con%rmed our notion that" for
a given initial load and average processing rates" there exist an
optimal L gain and an optimal balancing instant associated with
the one-shot L policy" which together minimi!e the average
overall completion time. This has also been veri%ed analytically
through our regeneration-theory-based mathematical model ./0.
1owever" this analysis has been limited to only two nodes and
has focused on handling an initial load without considering
subsequent arrivals of loads.
)n a existing static L policy" the scheduling decisions are
predetermined" while" in a dynamic load-balancing 2DL3 policy" the
scheduling decisions are made at runtime. Thus" a DL policy can be
made adaptive to changes in system parameters" such as the tra(c in
the channel and the unknown characteristics of the incoming loads.
Additionally" DL can be performed based on either local information
2pertaining to neighboring nodes3 " or global information" where
complete knowledge of the entire distributed system is needed before
an L action is executed.
DL 4 #xisting 'ystems are5.
63 The shortest-expected delay 2'#D3 policy
43 -ever-queue 2-73 policy &

-7 policy" all external loads are assigned to a
node that has an empty queue. )f more than one node
have an empty queue" the '#D policy is invoked among
the nodes with the empty queues to choose a receiver
node. 'imilarly" if none of the queues is empty" the '#D
policy is invoked again to choose the receiver node
among all the nodes.
which we have adapted to our distributed-computing
setting.
choose the receiver node among all the nodes. ,e implemented
the '#D and the -7 policies to perform the distributed computing
experiments on our test bed. The experiments were conducted
between two nodes connected over the )nternet 2keeping the same
processing speeds per task3. ,e performed three types of experiments
for each policy& 63 node 6 receiving" on average" 48 tasks at each
arrival and the average inter arrival time set to 64 s while no external
tasks were generated at node 4" 43 node 4 receiving" on average" 49
tasks at each arrival and the average inter arrival time set to : s" and
;3 node 6 and node 4 independently receiving" on average" 68 and 69
external tasks at each arrival and the average inter arrival times set to
: s and < s" respectively. #ach experiment was conducted for a two-
hour period.
Limitation of Existing System
 'cheduling decisions are predetermined
 L Threshold value is static
 'erver allotting process is very slow
 'ome time server would be maintaining all process
or each request the server will be allotted another
server.
 'erver would be switch over for every process" the
processing Time would be waste.
Proposed System
A sender-initiated distributed DL policy where each node
autonomously executes L at every external load arrival at that node.
The DL policy utili!es the optimal one-shot L strategy each time an
L episode is conducted" and it does not require synchroni!ation
among nodes. #very time an external load arrives at a node" only the
receiver node executes a locally optimal one-shot-L action" which
aims to minimi!e the average overall completion time.
This requires the generali!ation of the regeneration-theory-based
queuing model for the centrali!ed one-shot L. $urthermore" every L
action utili!es current system information that is updated during
runtime. Therefore" the DL policy adapts to the dynamic environment
of the distributed system.
Advantages of =roposed 'ystem
• The scheduling decisions are made at runtime
• L Threshold value is dynamic
• 'erver allotting process is dynamically predetermined
• Download process is very fast
Objectives
The expected value of the overall completion time for a given
initial load under the centrali!ed one-shot L policy for an arbitrary
number of nodes. The overall completion time is de%ned as the
maximum over completion times for all nodes. ,e use the theory to
optimi!e the selection of the L instant and the L gain. A distributed
and adaptive version of the one-shot is also developed and used to
propose a sender-initiated DL policy. Throughout the paper" a task is
the smallest 2indivisible3 unit of load and load is a collection of tasks.
ard!are Interface
 1ard disk & >8 ?
 @A* & 964 *
 =rocessor 'peed & ;.88?1!
 =rocessor & =entium )A =rocessor
Soft!are Interface
 BDC 6.9
 Bava 'wing
 *'-Access

Sponsor Documents

Or use your account on DocShare.tips

Hide

Forgot your password?

Or register your new account on DocShare.tips

Hide

Lost your password? Please enter your email address. You will receive a link to create a new password.

Back to log-in

Close