The fitness function. Each individual string will have a

The work was proposed to solve various dynamic
load-balancing problem using genetic algorithm 8a. Centralized Genetic
algorithm based method is used to perform load balancing operation where
sliding window technique is used to initialize a population of possible
solutions. The objective function such as maxspan, average utilization and
number of acceptable processor queues are used to obtain fitness function.
After completing roulette wheel method for selection, cyclic crossover and swap
mutation operation, the population of strings will be evaluated using the
fitness function. Each individual string will have a new fitness value and a
new probability of surviving into the next generation. The proposed dynamic
load-balancing approach developed using genetic algorithm minimize
interprocessor communication and total completion time, and provide better performance
by maximum processor utilization. The genetic-based load-balancing algorithm is
compared with the first-fit heuristic in terms of threshold policies,
information exchange criteria, interprocessor communication and other issues of
load balancing.

A method based on genetic algorithms for scheduling
and load balancing in parallel heterogeneous multi-processor systems was
introduced 1a. Permutations encoding is used to encode chromosomes. The
fitness function is calculated from the random chromosomes generated in the
initial population. Genetic Algorithm uses rotating roulette wheel strategy
selection. One point crossover is used in crossover operation and randomly
selecting two tasks and swapping them to perform mutation operation. The
simulations were performed using Matlab software for Genetic Algorithm, LPT,
SPT and FIFO. The result obtained is Genetic Algorithm provides better system
utilization with minimum response time compared to that of LPT, SPT and FIFO.
Genetic Algorithm for scheduling and load balancing can provide similar results
in different scales, and proves the robustness of the Genetic algorithm in
different scales.

We Will Write a Custom Essay Specifically
For You For Only $13.90/page!


order now

            The
proposed dynamic load balancing algorithm approach for heterogeneous
distributed system 2a. The sliding window technique is used to initiate a
population of possible solution. Centralized genetic algorithm perform load
balancing operation. LERT-MW method are used to schedule the job from and at
random interval of time genetic algorithm is applied. Selection, crossover and
mutation operation are performed, and fitness function is based on objective
function. The simulation is done using Matlab 6.0 where comparison of GA is
done with LERT-MWM algorithm. Total computation time reduces in GA compared to
that of LERT_MWM algorithm when the number of task increase with better
utilization of resource.

The proposed a genetic algorithm method 7a with
suitable fitness function for efficient sender-initiated load balancing in
distributed system. An overloaded processors (sender) initiate dynamic load
balancing by sending excess task to an under loaded processors (receiver).
Genetic operators are applied to the population of a processor in distributed
system. Sender send the request message to the processors. After receiving the
request message, processors send back accept message or reject message based on
its own CPU queue length. Genetic algorithm is used to decide suitable
receivers based on its fitness value, to which the request message should be
sent. When a sender receivers accept message from a processor, then it send a
task to that processor. Several simulation is done to compare proposed method
with conventional sender-initiated algorithm. The results shows the
effectiveness of the proposed algorithm, since the performance of the proposed
scheme is better than that of the conventional scheme on the response time and
mean response time.