A Seminar Report ON Job Scheduling in Multi Processor Architecture Using Genetic Algorithm

You might also like

Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 21

A

SEMINAR REPORT
ON
JOB SCHEDULING IN MULTI PROCESSOR ARCHITECTURE USING GENETIC ALGORITHM

SUBMITTED TO: MUNISH KHANNA SIR

SUBMITTED BY: DEEPAK KUSHWAHA(3rd Year)


KAMLESH KUMAR(3rd Year)
NIKHIL GUPTA (3rd Year)

(DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING)


JOB SCHEDULING
Job scheduling is an important issue which has many applications in
different fields.
The main issue is how jobs are partitioned between processors in
which total finishing time and waiting time are minimized.
Minimization of these two criteria simultaneously, is
a multi objective optimization problem.
To solve this problem, a genetic algorithm is proposed to optimize
two objective simultaneously.
In so doing, fitness function based on aggregation is used .
In addition, longest processing time and shortest processing time
algorithms are implemented to compare with genetic algorithm.
Proposed genetic algorithm shows better results in experiments and
can reduce finishing time and waiting time simultaneously.
PROPOSEDMODEL
GA is an evolutionary algorithm that increasingly has been used in many fields of
artificial intelligence.
Main idea of GA is based on natural genetic rules.
PRESENTING SCHEDULE
Population of chromosomes is used by GA to represent solution.
 Population is set of chromosomes and chromosomes is set of genes.
Therefore, solutions should be represented in the form of genes which is called
gene coding.
When using GA for scheduling problems, many scholars used integer encoding.
which provides a convenient and natural way to express the mapping from
representation to solution domain.
The solution to job partition is encoded into a vector of integers, where the ith
element of an individual is k if job assigned to processor k.
POPULATION INITIALITION
The population initialization creates the first generation of
population and determines the starting point of the GA search
routine.
Number of processors, population size and number of jobs are
needed to genetic initial population.
Number of processors and population size are entered by user.
Required time slice of each job are read from input file.
Initial population of chromosomes is generated by using these
parameters.
Each chromosomes of initial population can be shown by an array
of integer numbers between 1 and number of processors.
These numbers present a random partition of jobs that makes a
random schedule
FITNESS FUNCTION
GENETIC OPERATORS
GA uses selection operator to select the superior and eliminate the inferior.
 The individuals are selected according to their fitness-the more suitable the
more chances they have to reproduce.
This operator generates next generation by selecting best chromosomes from
parents and offspring.
Objective function can be used as fitness function.
 Chromosomes can bed by using their fitness values.
Although GA does not search all of search space to find answer, smaller search
space can make GA to find better answers in less time.
 A heuristic is used to reduce search spaces . In so doing , jobs of chromosome
were sorted by using SPT method before calculating fitness.
For the scheduling problem, the double-point crossover
operater is used: two crossover points are generated
uniformly in the mated parents at random, and then the
two parents exchange the middle portion between these
crossover points to create the new offspring(Fig. 1).
EXPERIMENTAL RESULTS
 Table 1 shows experimental results for GA ,
LPT and SPT algorithms together with total
finishing time (FT) and waiting time (WT). It can
be seen that GA can reduce both of WT and FT
but LPT and SPT improve only one of them.
CONCLUSION
Scheduling problem is a NP-Hard problem. Therefore, uncertain
methods such as evolutionary algorithms and specially GA are
suitable in this case.
The goal is to partition jobs between processors such that
minimum FT and WT is achieved.
Simultaneously minimizing these objectives is a multi objective
optimization problem.

Simulation is done by using SPT, LPT and GA with fitness function


based on aggregation. Experimental results show that GA can
reduce both of FT and WT criteria simultaneously but LPT and SPT
can reduce only one of them.
Mathematical model:
1.

2.

3.

4.
WHO DOMINATE:

>> Same is the case with objective vectors.


SUPERIORITY:
In general there are three types of fitness function:
1. Aggregation based.
2. Criteria-based.
3. Based on pareto-dominance.

An aggregation based approach:


Problem definition:

>> JSP is a NP-hard problem.

Let we have M different processors, and N different jobs:


Then,
>> This is a biobjective paper:

>> our main job is to minimise waiting time and

finishing time.
>> Two objectives can be unified using the waited

sum:

f = total finishing time+ß waiting time.

ß is the weight coefficient , which can be adjusted


to compromise between two objectives.

You might also like