SECTION I.

Introduction
Last Few years, Cloud is emerging technology which satisfies for different types of users in
different ways. Basically all requirements are handled by cloud with different deployment
model- SAAS, PAAS, IAAS, KAAS, DAAS, LAAS & etc. Clouds works on use and pay
mechanism that provides loyal services through CCs (cloud centers). Cloud is coming after
of distributed computing which has gives extra facility like Virtualization. Consumer can
enlarge the computing power and storage capacity throw elasticity. Cloud computing is
build to use applications that only place on different location and also provide different type
of virtualized platform that accomplish cloud consumers jobs in minimum time and
minimum cost. Many job scheduling algorithm have been proposed for handling job in
smart way by researcher [1]–[2][3].

There are some evaluation parameters which decide a job place on smart way like
performance, throughput, utilization of resources, response time, make span, load balancing
index & etc. Some scheduling algorithms are focused on QOS parameters also. Traditional
scheduling algorithms are not provides good result in cloud but have some issue. Scheduling
algorithm is classified in two ways: Batch mode & Online mode. In first, jobs are collected in
fixed set based on arrival in cloud. First Come First Serve, Shortest Job First, Round Robin,
Min-Min, Max-Min, RASA all are techniques used in batch mode scheduling. In online
mode, at time of arrival of jobs, are scheduled. Most Fit Task Scheduling (MFTS) is an
example of online mode [1].

Scheduling of jobs based on priority is a concern issue because every job wants to execute in
limited time period. Some job scheduling algorithm must consider priority [4]. For resolve
this problem there are lots of algorithm consider priority of jobs [2]–[3][4][5][6][7][8] [12].

For solving this issue Combinatorial Optimization algorithm can be used. One example of
this algorithm is Hungarian Method. Hungarian Method gives solutions of assignment
problem in polynomial time. This method is developed in 1957 by two great Hungarian
mathematicians: Denes Konig and Jeno Egrary can be used in cloud computing to achieve
better result in scheduling process.

The main objective of this paper is mainly focused on priority based scheduling. Proposed
algorithm gives improvement in completion time and consistency for allocating jobs. The
rest of this paper is organized by following section as: section 2explain Hungarian Method
with calculations, section 3 Proposed scheduling algorithm based on Hungarian on cloud
environment, section 4 Experimental setup for analyzing algorithm finally section
5 & 6 results and conclusion of this paper.

SECTION II.
Hungarian Method
Assignment of jobs is a crucial things but Hungarian method [11] gives best solution without
any direct or indirect comparison of each and every solution. It works on principle of
minimizing the given cost matrix to a opportunity costs matrix. Opportunity costs shows the
relevant penalties associated with assigning resource to an activity as opposed to making the
better or minimum cost assignment. If we can able to minimize the cost matrix to the extent
of having at least one zero in each column and row, then it might be possible to make better
assignments (opportunity costs are all zero). Hungarian method is a method that works on
combinational optimization algorithm which solves the assignment problem in polynomial
time.

Steps of Hungarian Method:

1. Find smallest value from each row and subtract that value from all the entries of its
row.

2. Find smallest value from each column and subtract that value from all the entries of
its column.

3. Draw lines across rows and columns so all the zero values of cost matrix are covered
and also minimum number of lines are used.

4. Testing Optimality: (I) If minimum number of covering lines is n then optimal

assignment of zeroes are possible and algorithm finishes. (ii) If minimum number of
covering lines is less than n then optimal assignment of zeroes is not possible. So
proceed to step 65.

5. Find the smallest value not covered by any horizontal or vertical line. Add that value
to each uncovered column, and subtract smallest value from each covered row.
Return to step 3.

Mathematical Representation:-

1. Define smallest value in each raw.

Table 1

2. Subtract smallest value from each row.

 Table 2

3. Subtract smallest value from that column.

Table 3

4. Count Minimum no of rows and column that covers all 7ro

Table 4

5. Test the Optimality Npo of lines =3<4 where 4=n Here, optimal scenario is not
satisfied so we forward to step 6.
6. from table 4 those values that are not covered in lines, find minimum value from
those available values & add that minimum value to intersection points, subtract that
minimum value from availed values.

Table 5

7. repeat step 4 for checking optimality condition.

Table 6
 8. from table 6 total no of lines covering zero is 3 which is not equal to 4, so repeat step
6 & 7.

Table 7

Hence number of lines are 4 now so we bet the optimal solution now we are going to allocate
tasks on VM. Job 1 is allocated on Vm 1, job 2 is allocated on Vm 4, job 3 is allocated on Vm
3 and last job is allocated on Vm 1.

SECTION III.
Proposed Algorithm
The proposed algorithm consists of two levels of priorities: Job level (based on Attribute) &
Resource Level (base on requirement). In Job level we define certain parameters like length
of job, requirement of bandwidth for execution of job, requirement of processing speed & at
Resource level we focused on how much instruction handled by resources (MIPS) & what is
the transfer rate of resources (MbPS). So this all parameter is used to calculate priority
based on Hungarian Method for achieving good performances and minimum cost.

1. Enter T={T1, T2, T3…Ti} set of Tasks/Jobs.

2. Enter R={R1, R2, R3…Ri} set of Resources.
3. Calculate priority of each jobs based on Hungarian Method.

4. Same Calculate priority of resources based on Hungarian Method.

5. Rearrange the list of jobs and Resources based on Priority.

6. Now bind the jobs of highest priority on updated resourc step 5: Test the optimality

7. Update list of jobs & Resources.

8. Check Allocation Completed

1. Yes then End.

2. No then follow the step 3 to 7.

SECTION IV.
Evaluation Parameters
This section focused on makespan and consistency which we consider to prove the
performance of proposed algorithm.
A. Completion Time
 This algorithm is chiefly concentrated on finding priority of tasks (jobs) and
resources, so better time complexity cannot be expected. Though this algorithm is
primarily concentrated on finding priority of tasks (jobs) and resources, proposed
method has tolerable completion time (makespan). The experimental result shows
this algorithm has tolerable completion time.

B. Consistency
 Consistency means each and all matrixes have logically reasonable value (which
is <0.1). It primarily rest on decision maker and can make a number of stable
matrixes based on Hungarian Calculation. Among all available methods such as
Eigen-value method, consistency of all matrixes can be calculated. In proposed
method, consistency check is not an issue as there is no need to check whether the
matrix is consistent or not. Hungarian matrix is consistent matrix.

SECTION V.
Experimental Results
Job scheduling based on Hungarian method shows healthier and adequate results when
compared to the traditional algorithm. The proposed algorithm focused on those jobs who
has highest priority for accomplish. For evaluation purpose, we used cloudsim resource
simulator which is a toolkit for modeling and simulation of cloud computing environment
and evaluation of resource provisioning algorithms [10]. Cloudsim simulator is used for
checking the execution and allocation of processing elements [PE] on cloud jobs. When jobs
are entered in to the cloud environment, proposed algorithm start to execute the jobs. For
that first calculation are done based on Hungarian method and get the priority of the jobs
based on requirement. Same calculations are done on resource side also. Then highest
priority jobs are allocated on best resource. Once the allocation is completed and execution
is completed the cloudsim automatically update the set of jobs and resource as well. For
check the proposed algorithm we take 100 jobs having large data volume and perform
proposed algorithm on different set of Resources which gives superior results.

Fig. 1.
Results Based on Different Parameter

View All

SECTION VI.
Conclusion
In cloud, scheduling of jobs based on requirement is a key area to think. There are so many
algorithms available for priority based scheduling. Hungarian method is different from
those traditional algorithms in terms of time as well as complexity. We proposed this novel
and distinct method that finds priority of jobs and resources that gives best performance.
This algorithm needs further improvement by considering large number of resources.