Job Scheduling in Grid Computing Using Ant Colony Optimization

You might also like

Download as ppt, pdf, or txt
Download as ppt, pdf, or txt
You are on page 1of 19

JOB SCHEDULING IN GRID

COMPUTING USING ANT


COLONY OPTIMIZATION

GUIDED BY :Ms.P.DEVAKI
TEAM MEMBERS:
N.DIVYA(07BCS13)
S.KAVIYA(07BCS24)
P.NATHIYA(07BCS29)
AGENDA
• What is Grid Computing?
• What is Job scheduling?
• Problem Statement
• Ant colony Optimization
GRID COMPUTING
 Powerful and innovative system for a group of
heterogeneous distributed computing systems.

 Denotes a dynamically changing set of


computing resources distributed among different
administrative domains.
JOB SCHEDULING
 We are given n jobs J1, J2... Jn of varying sizes,
which need to be scheduled on m heterogeneous
machines, (n>m) with the minimum Makespan.
PROBLEM STATEMENT
 We need high performance computing systems for data
and computational intensive problems.
 Grid has evolved as an utility computing and user can
submit their jobs to grid to be solved quickly.
 Stagnation in grid may occur when all jobs require same
resources.
 So, Job Scheduling in grid computing is the major issue
for the researchers.
 In order to overcome this problem, our project aims to
use Ant Colony Optimization Algorithm.
SYSTEM ARCHITECTURE
ANTCOLONY OPTIMIZATION
 A colony of ants work together to find the
shortest path between their nest and food source.
 Every ant will deposit a chemical substance
called pheromone.
 They choose the shortest or optimal path based
on the pheromone value.
 The Algorithm focuses on local pheromone trail
update and the trail limit values to minimize the
makespan.
INPUT PARAMETERS:
ETC Matrix
No.of Tasks-n (0<n<512)
TASK REQUIREMENTS
 Task requirements matrix represents the user
requirements of QoS factors for executing the
particular task.
 TaskRequirementsik =
MACHINE CAPABILITY
• Machine capability for each machine indicates the
QoS factors(cost,RAM,Deadline)associated with
machine j for each task i.
• The machine capability matrix is:

• This matrix is calculated based on the following


formula:
Ti*Mj={Cost(j)*etc(ij), RAM, etc(ij)}
ASSUMPTIONS:
 we have assign a guided probability value for
each QoS factor(Cost,RAM,Deadline).
WQoS=(0.5,0.25,0.25).
SATISFY OPERATOR

• k = the number of QoS parameters.

• Rj₪ Ti means that the resource Rj can satisfy


the task Ti and guarantees QoS parameters.
INITIAL PI CALCULATION
 If all the QoS factors are satisfied then, the initial
pheromone indicator value is calculated based on
the following equation:
INITIAL PI CALCULATION-
contd
 Hence for m tasks and n machines, PI matrix
will be of the form:
M1 M2 …. Mm
PI= T1 PI 11 PI 12 ..... PI im
.. ... ..... ……
Tn PI n1 PI n2 ..... PInm
SCHEDULING OF TASKS
 In each iteration we select the largest entry from the
matrix PI.
 For instance if PI 11 is the largest entry in the matrix

 If no other taski is allocated to the machine1 already,


then the task i is allocated to machine j.
 Else the task is allocated to the next machine that
has the next highest pheromone value.
LOCAL UPDATE
 The local pheromone update is given by the
formula:
 PIij = (1-ρ) PIij + ρ.1/Nr
Where,
PIij -> PI values on the machine j where
the task is allocated.
ρ -> evaporation rate of the pheromone
(0< ρ<1) .
Nr ->The number of resources or machines in the
grid.
GOLBAL UPDATE
 After all the machines have been assigned a task
global update is made to the entire PI matrix.
 Global update performed on the PI matrix based
on the following formula:
PImodified =
(1- ρ ) PIinitial * 1/MachineAvail[i]
OUTPUT
• 512 tasks scheduled across 16 machines.
EXPERIMENTAL RESULTS
TAS
K/
ALG
MIN-MIN MAX-MIN ANT COLONY
CONSIST INCONSI PARTIA CONSIS INCONSI PARTIAL CONSIS INCONS PARTIA
ENT STENT L TENT STENT CONSIST TENT ISTENT LCONSI
CONSIS ENT STENT
TENT

250 419.15 195.79 406.60 425.51 184.61 402.55 376.50 359.08 371.68
300 516.84 236.52 496.04 523.08 225.49 495.44 369.69 355.64 368.40
350 556.75 257.86 536.02 536.21 247.95 535.16 374.52 348.21 373.63
400 633.85 291.86 614.50 639.73 283.59 610.55 373.94 348.18 386.00

You might also like