618 The International Arab Journal of Information Technology, Vol. 20, No.

4, July 2023

Exact Algorithm for Batch Scheduling on

Unrelated Machine
Hemmak Allaoua
Department of Computer Science, Mohamed Boudiaf University,
Laboratory of Informatics and its Applications of M'sila, Algeria
[email protected]

Abstract: In this paper, we propose a new linear algorithm to tackle a specific class of unrelated machine scheduling
problem, considered as an important real-life situation, which we called Batch Scheduling on Unrelated Machine (BSUM),
where we have to schedule a batch of identical and non-preemptive jobs on unrelated parallel machines. The objective is to
minimize the makespan (Cmax) of the whole schedule. For this, a mathematical formulation is made and a lower bound is
computed based on the potential properties of the problem in order to reduce the search space size and thus accelerate the
algorithm. Another property is also deducted to design our algorithm that solves this problem. The latter is considered as a
particular case of RmCmax family problems known as strongly NP-hard, therefore, a polynomial reduction should realize a
significant efficiency to treat them. As we will show, Batch BSUM is omnipresent in several kind of applications as
manufacturing, transportation, logistic and routing. It is of major importance in several company activities. The problem
complexity and the optimality of the algorithm are reported, proven and discussed.

Keywords: Scheduling, unrelated machine, exact method, parallel machine, batch scheduling.

Received September 9, 2022; accepted May 4, 2023

https://doi.org/10.34028/iajit/20/4/8

1. Introduction to tackle NP-complet scheduling problems under

“divide and conquer” paradigm. Two potential
In most cases, unrelated machine scheduling without
properties of the problem will be stated then proven that
preemption is considered among the hardest scheduling
allow to elaborate then prove our algorithm. The first
problems in the strong sense regarding their
property allows to find a lower bound of the optimal
complicatedness and the astronomic number of
solution that is to reduce the search space size and thus
candidate solutions especially when the number of jobs
the processing time of the algorithm. The second one
and machines are large enough. In the other hand, these
leads to compute the optimal solution of the problem.
kinds of problems are omnipresent in many company
In the rest of this paper, a related work is presented
activities such as manufacturing, transportation, logistic
in section two, then we give a full description of the
and routing. Indeed, finding optimized machine
treated problem in section three. Section four is
schedules is an important and challenging task, as a
dedicated to expose the design, phases and complexity
large number of jobs need to be processed every day.
of our algorithm. Some comments and discussion are
They reveal a significant impact on the income of the
reported in section five.
company, especially when we need to minimize the last
completion time known in scheduling field as
makespan (Cmax). The problem treated in this work 2. Related Work
consists on a specific case of the class of problems A significant amount of research on scheduling
denoted as Rm| |Cmax which is proved as strongly NP- problems in general, and on those of unrelated
hard [5]. This case consists to schedule a batch of machines has been studied extensively. Scheduling
identical and non-preemptive jobs on unrelated identical jobs on unrelated machines have been the
machines which we have called Batch Scheduling on subject of thorough research in the past, and two
Unrelated Machine (BSUM) and we denoted as surveys by Allahverdi et al. [6] and Allahverdi [7] give
Rm|pij=pj|Cmax. Its resolution consists to find the an overview of the related literature. Relevant research
optimal vector of job numbers to be assigned to the in this type of problems includes approximation
machines that minimize the makespan. A new linear algorithms [4, 6, 8, 29] exact algorithms [16, 17, 20],
algorithm is proposed, proven then implemented to mathematical programming techniques [14, 22, 25]
tackle this problem. We aim to provide a tool that optimization techniques [4, 8, 9, 23], and metaheuristic
determine a polynomial reduction for the Rm||Cmax approaches [1, 18]. Mokotoff [21] and Pinedo [26]
family problems to Rm|pij=pj|Cmax that is to contribute provides an extended survey for multiprocessor jobs
problems in general.
Exact Algorithm for Batch Scheduling on Unrelated Machine 619

The problems of scheduling on unrelated machines For the instance above, there are 21 different
to minimize the makespan were also been well studied manners to dispatch the 5 jobs between 3 machines
in the literature [5, 12, 19, 27]. Since this class of shown with their respective makespan (Table 2):
scheduling problems is known and proved as strongly
Table 2. Exhaustive list of solutions.
NP-hard, all these works gave approximate algorithms
[3] to solve them. in some work, just the case of two (5,0,0) 25 (1,4,0) 40 (4,1,0) 20
types of jobs is considered, Vakhania et al. [28] (0,5,0) 50 (1,0,4) 32 (4,0,1) 20
Hernandez presented a polynomial time algorithm. (0,0,5) 40 (0,1,4) 32 (0,4,1) 40
Ebenlendr et al. [10] elaborate a O(n2)-algorithm to (2,3,0) 30 (3,2,0) 20 (1,1,3) 24
tackle a special case of the class Rm||Cmax [2, 21]. (2,0,3) 24 (3,0,2) 16 (1,3,1) 30
Fanjul-Peyro and Ruiz [13]. Some research focus on the (0,2,3) 24 (0,3,2) 30 (3,1,1) 15
equal processing times of jobs [11, 15]. Munir et al. (2,1,2) 16 (2,2,1) 20 (1,2,2) 20
[24] propose novel approaches for Scheduling task
graphs in heterogeneous distributed computing The optimal solution is (3, 1, 1) with the makespan
environment that tackle a similar problem. Cmax=15 represented in the diagram below (Figure 1):

3. Problem Definition and Overview Machine 1 5 5 5

Machine 2 10
3.1. Problem Statement Machine 3 8
A batch of n identical and non-preemptive jobs to be Figure 1. Schedule diagram.
scheduled on m unrelated parallel machines. The
processing time of one job of the batch on the machine j Construct a solution to this problem consists to
is pj (assuming that pj is integer and pj>0); pj is the time dispatch n jobs one by one between m machines, that
spent by the machine j to proceed one job of the batch leads to separate a sequence of n “1”s with (m-1) “,”s to
without preemption. That show that the speed of the form m subsequences then add the “1”s of each
machine j is inversely proportional to the processing subsequence.
time pj. We aim to find the schedule of these n jobs Example: for n=5 and m=3 the sequence 111, 11
with minimum last completion time (makespan) Cmax. define the solution (3,0,2); that means we assign 3 jobs
This scheduling problem can be denoted as to the first machine, any job to the second machine and
Rm|pij=pj|Cmax. 2 jobs to the third machine and so on.
Below (Table 1) an instance of BSUM problem: Thus, the number of ways to dispatch n jobs between
m machines equals to the number of ways to separate n
Table 1. An instance of BSUM.
“1”s by (m-1) “,”s ; therefore, the number of candidate
n = 5 jobs ; m = 3 machines 𝑚−1 (𝑛+𝑚−1)!
Machine j 1 2 3 solutions is: 𝐶𝑛+𝑚−1 = 𝑛!×(𝑚−1)! , that explodes when
Processing time pj 5 10 8
n is big enough.
Since, in unrelated machine scheduling, each  Example
machine has its own speed vj=pij/pi, but in our case all
With n=5 jobs and m=3 machines, we have 𝐶72 =
jobs are identical, so the processing time pj determine
the speed of the machine j. 21 different ways to dispatch 5 jobs on 3 machines as
given above. For 100 jobs and 20 machines, the search
 Problem Formulation space size becomes astronomic: 491037×1016 candidate
𝑚−1
The resolution of this problem consists to dispatch the n solutions. Result: since 𝐶𝑛+𝑚−1 ≈ 𝑂(2𝑛 ), exhaustive
jobs on the m machines such that Cmax is minimal. algorithm still inefficient. It is why we have to look for
Therefore, the biggest task to do is to determine the a polynomial algorithm based on potential properties of
number xj of jobs to be assigned to the machine j for the problem BSUM to solve it.
j=1, m; thus, the solution of this problem can be
represented as an integer vector x=(xj)j=1,m that describe 3.2. Problem Properties
this assignment. Hence the problem can be formulated  Property 1
as below:
𝑛
min𝑥∈𝑁𝑚 𝐶𝑚𝑎𝑥 = max(𝐶𝑗 ) = max(𝑥𝑗 𝑝𝑗 ) ; 𝐿𝐵 = ⌈ 1 ⌉ is a lower bound of Cmax for BSUM.
∑𝑚
𝑗=1 𝑝𝑗
𝑥 = (𝑥𝑗 ) ; 𝑗 = 1, 𝑚
𝑠. 𝑡. 𝑥𝑗 𝑝𝑗 ≤ 𝐶𝑚𝑎𝑥 ; 𝑥𝑗  𝑁 ; 𝑗 = 1, 𝑚
 Proof
{ ∑𝑚𝑗=1 𝑥𝑗 = 𝑛
From the constraints in the BSUM formulation, we
Exhaustive research of the optimal solution amounts to 𝐶𝑚𝑎𝑥 𝑚 𝐶𝑚𝑎𝑥
seek all ways to partition the integer n as an have: 𝑥𝑗 ≤ 𝑝 . Hence: ∑𝑚 𝑗=1 𝑥𝑗 ≤ ∑𝑗=1 𝑝
𝑗 𝑗
arrangement of m integers whose sum is n.
620 The International Arab Journal of Information Technology, Vol. 20, No. 4, July 2023

1
Thus: 𝑛 ≤ 𝐶𝑚𝑎𝑥 × ∑𝑚
𝑗=1
(1) implies that the search start from LB, the property
𝑝𝑗
𝑛
(2) implies that we have to look for the smallest
Therefore: 𝐶𝑚𝑎𝑥 ≥ 1 multiple of one the processing time pj that is the last
∑𝑚
𝑗=1 𝑝𝑗
completion time in the schedule.
Since Cmax is integer: Cmax≥LB
That means: 4. Algorithm Description
𝑚
𝑛 1
1
; 𝑖𝑓𝑛 𝑚𝑜𝑑𝑢𝑙𝑜 ∑ = 0
𝑝𝑗 In this section we will describe and discuss all phases
∑𝑚
𝑗=1 𝑝 𝑗=1
𝑗 of our proposed linear algorithm for solving BSUM.
𝐿𝐵 =
𝑛 This approach consists of three phases:
(𝑖. 𝑒. 𝑖𝑓 ∈ 𝑁) ;
1 The first phase is computing lower bound of the
∑𝑚
𝑗=1 𝑝
𝑗
𝑛 makespan. Based on the property (1) above we
+ 1 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 ;
1 elaborate the Algorithm (1) below:
∑𝑚
𝑗=1 𝑝
{ 𝑗

Algorithm 1: Int LB(int n ; int m ; int[] p)

5
For the instance above: 𝐿𝐵 = ⌈1 1 1⌉ = 12. # Computing the lower bound LB.
+ +
5 10 8 # Input: number of jobs n, number of machines m and the
 Property 2 respective processing time table p[].
# Output : LB.
The minimal value of Cmax for BSUM is the smallest {
multiple  of one of the integers pj that satisfy: sum = 1 / p[0]
 for ( j = 1 to m-1)
∑𝑚
𝑗=1 ⌊𝑝 ⌋ ≥ 𝑛. sum = sum+ 1 / p[j]
𝑗
If ( n mod s = 0 )
 Proof Return (n / sum)
else
First, we have to prove that  exists. So, for a given return LB=((int)(n / sum) + 1)
instance (n;m;pj,j=1,m), there exists at least the }
makespan: n ×minj=1,m(pj) which is a multiple of one of It is clear that this algorithm is linear, in O(m).

the pj integers and satisfy ∑𝑚
𝑗=1 ⌊𝑝 ⌋ ≥ 𝑛 , that is when The second step consists to compute the makespan
𝑗
based on the property (2) of BSUM. The main idea is as
we assign all jobs to the fastest machine where the follow: starting from the LB computed in the Algorithm
solution is in the form (0,….,0,n,0,..,0), therefore,  (1) above, we seek progressively for the multiple of the
exists. Assume that xj is the number of jobs to be integers pj that satisfy the property (2), whence the
proceeded by the machine j, the completion time Cj of Algorithm (2) below:
the machine j is then Cj =xj ×pj. Since the makespan of
the schedule is defined as Cmax=maxj=1,m(Cj); Thus: Algorithm 2: Int MinCmax(Int n ; Int m ; Int[] p)
∃ 𝑖 ∈ [1, 𝑚] ∶  = 𝑥 × 𝑝𝑖 ; 𝑥 ∈ 𝑁 ∗ . # Computing the makespan min Cmax.
( is one among the Cj , 𝑗 = 1, 𝑚). # Input: number of jobs n, number of machines m and the
respective processing time table p[].
So,  is a multiple of one of the integers pj. # Output: min Cmax.
Since  is a makespan of the schedule then: {

∀𝑗 ∈ [1, 𝑚]:  ≥ 𝑥𝑗 × 𝑝𝑗 , Hence: ∀𝑗 ∈ [1, 𝑚]: ≥ 𝑥𝑗 Cmax = LB( n , m , p[]) ; # call LB function
𝑝𝑗 SumQ = 0 # Sum of Quotients
  while (True)
By adding: ∑𝑚
𝑗=1 ⌊𝑝 ⌋ ≥ ∑𝑚
𝑗=1 𝑥𝑗 ; ∑𝑚
𝑗=1 ⌊𝑝 ⌋ ≥𝑛
𝑗 𝑗 {
In the other hand =minj=1,m(Cmax), so  is the for (j = 0 to m-1)
if (Cmax Mod p[j]) = 0) break # exit for
smallest multiple of one the integers pj that satisfy:
if (j <= m)

∑𝑚
𝑗=1 ⌊𝑝 ⌋ ≥ 𝑛. {
𝑗 SumQ = Cmax / p[0]
From these two properties, we deduct the following for ( j = 1 to m-1)
corollary: SumQ = SumQ + Cmax / p[j]
if (SumQuotients >= n)
 Corollary Return MinCmax = Cmax # exit while
}
The minimal value of Cmax for BSUM is the smallest
Cmax = Cmax +1
multiple  of one the integers pj that satisfy: }
 𝑛
𝑗=1 ⌊𝑝 ⌋ ≥ 𝑛 and  ≥
∑𝑚 1.
}
𝑗 ∑𝑚
𝑗=1 𝑝𝑗

These two properties allow to find the makespan and

the optimal solution in polynomial time. The property
Exact Algorithm for Batch Scheduling on Unrelated Machine 621

𝐿𝐵 generated. The algorithm has been implemented in the

Once the LB is computed, we can assign 𝑥𝑗0 = ⌊ ⌋
𝑝 𝑗 C programming language and compiled with gcc
jobs to the machine j to construct the initial solution version 4.8.2. The computational experiments have
𝑥 0 =(𝑥𝑗0 ) where ∑𝑚 0
𝑗=1 𝑥𝑗 ≤ 𝑛. been performed on one core of a system with Intel Core
If ( ∑𝑚 0 0
𝑗=1 𝑥𝑗 = 𝑛 ) then 𝑥 is the optimal solution
i5-4210U processor at 1.7 GHz and 10 GB of RAM
and LB =min Cmax. under a Linux OS.
If ( ∑𝑚 0 0
𝑗=1 𝑥𝑗 < 𝑛 ) then 𝑥 is not feasible solution
and LB < min Cmax, we will have to assign the
remaining jobs not yet assigned whose the number is: =
𝐿𝐵
𝑛 − ∑𝑚
𝑗=1 ⌊ 𝑝 ⌋ . As upper bound, we can assign them
𝑗
to the fastest machine (i.e., the machine with min(pj)),
so there exists a feasible solution x with:
Cmax=LB+r×min(pi). (x may be not optimal).
Therefore, this algorithm terminates and converges
because  exists as proven in property (2). Figure 2. Implementation interface.
The while loop makes at most r iterations as much as
the two inner successive loops are in O(m). Therefore, In order to show the efficiency and the robustness of
In the worst case, the Algorithm (2) is in O(rm) (that is our algorithm a set of random input data is generated
when Cmax= Cmax=LB+r×minj=1,m(pi)). using our own random generator that is to run the
By replacing LB by its value in the expression of r, algorithms with same input. The following
we find r≈ 𝑚. Since, in practice m<<n, say m ≈ c. 𝑛 experimental settings is used:
(c<1) this algorithm is at least in O( n). # of jobs n {50,100,500,1000,10000,100000};
The third step consists to find the optimal solution # of machines m {n/20 , n/10 , n/5};
using the second part of the property (2), the optimal instance k  [1,10] .
solution is a vector x=(xj)j=1,m that describe the Processing time pj: random integer in the range [1,
assignment of the n jobs to the m machines; that is done 20]. Following an example of results fo 10 instances for
by dividing Cmax respectively by the processing times the set n=1000 and m=100 (Table 3).
pj, that means assigning the n jobs to the m machines
Table 3. results for n=1000 and m=100.
one by one. The number of jobs assigned to the
𝐶 instance Time (ms) Cmax
machine j is 𝑥𝑗 = 𝑚𝑎𝑥 (the quotient of Cmax by pj), as 1 76 70
𝑝𝑗
2 65 72
shown in the Algorithm (3) below: 3 73 52
4 74 72
Algorithm 3: Int[] Solution (Int n; Int m; Int[] p; Int Cmax ;)
5 75 60
#Finding an optimal solution. 6 76 52
#Input: number of jobs n, number of machines m, the 7 86 57
8 71 60
respective processing time table p[]. 9 71 54
#Output: schedule of jobs (x m-vector of jobs number 10 72 72
assigned to the m machines) and Cmax. Average 73.9
{
Int Sol (m) The algorithm was run for all the data set, then we
Int AllJobs = 0 have constructed the curve representing the CPU time
for (machine = 0 to m-1)
{ average in terms of n for each case of m values (Figure
Sol [machine]=0 3).
for (Jobs=1 to Cmax / p[machine])
{
Sol[machine]++
AllJobs = AllJobs+1
if (AllJobs = n) return Solution = Sol
}
}
}
In order to show the algorithm efficiency, we have
implemented it with the interface shown in the figure
bellow (Figure 2). The data instances are generated
randomly, that allow us to introduce instances with big Figure 3. Average CPU time in terms of n.
size (large number of jobs and/or machines). For each
case of these two dimensions, ten instances are This curve shows clearly the linearity of the
622 The International Arab Journal of Information Technology, Vol. 20, No. 4, July 2023

algorithm complexity whatever the choice of m. [2] Afzalirad M. and Rezaeian J., “A Realistic
Variant of Bi-Objective Unrelated Parallel
 Comparison with other Exact Approaches
Machine Scheduling Problem: NSGA-II and
As we are about to discuss exact approaches, where the MOACO Approaches,” Applied Soft Computing,
optimality must be formally proven, we have compared vol. 50, pp. 109-123, 2017.
our algorithm to the exact ones elaborated for the same https://doi.org/10.1016/j.asoc.2016.10.039
problem found in the literature. the results are [3] Afzalirad M. and Shafipour M., “Design of An
summarized in the Table 3 below. All these cases are Efficient Genetic Algorithm for Resource-
reported in several papers and formally proven. Some Constrained Unrelated Parallel Machine
of them was served to measure and justify the Scheduling Problem with Machine Eligibility
efficiency of heuristics [5, 7]. Restrictions,” Journal of Intelligent
Table 3. Comparison with other algorithms. Manufacturing, vol. 29, no. 2, pp. 423-437, 2018.
Approach Complexity
DOI:10.1007/s10845-015-1117-6
Linear Assignment O(mn2) [4] Afzalirad M. and Rezaeian J., “Resource-
Dynamic Programming O(mn2m + 1) Constrained Unrelated Parallel Machine
Integer Linear Programming O(n log m)
Linear programming relaxation O(n +mlogm)
Scheduling Problem With Sequence Dependent
Setup Times, Precedence Constraints and
Note that, the least expensive metaheuristic as Machine Eligibility Restrictions,” Computers and
simulated annealing will make not less than O(n2) time Industrial Engineering, vol. 98, pp. 40-52, 2016.
to give just a good approximate solution (the number of https://doi.org/10.1016/j.cie.2016.05.020
iteration must at least be linear in n and the [5] Allahverdi A., “The Third Comprehensive
computation of a solution neighbour costs O(n) . Survey on Scheduling Problems with Setup
Times/Costs,” European Journal of Operational
Research, vol. 246, no. 2, pp. 345-378, 2015.
5. Conclusions
https://doi.org/10.1016/j.ejor.2015.04.004
In this paper, a new algorithm was proposed then [6] Allahverdi A., Ng C., Cheng T., Kovalyov M.,
implemented for solving a specific class of unrelated “A Survey of Scheduling Problems with Setup
machine scheduling problem where we have to Times or Costs,” European Journal of
schedule a batch of same jobs on unrelated machines Operational Research, vol. 187, no. 3, pp. 985-
which we have called BSUM. The algorithm is 1032, 2008.
designed based on the potential properties of the https://doi.org/10.1016/j.ejor.2006.06.060
problem. We showed that this algorithm is quadratic [7] Allahverdi A., “The Third Comprehensive
complexity in worse case. For this, a mathematical Survey on Scheduling Problems with Setup
formulation is made and a lower bound is computed Times/Costs,” European Journal of Operational
based on the potential properties of the problem in Research, vol. 246, no. 2, pp. 345-378, 2015.
order to reduce the search space size and thus accelerate https://doi.org/10.1016/j.ejor.2015.04.004
the algorithm. Another property is also deducted to [8] Chen Z., “Parallel Machine Scheduling with
design our algorithm that solves this problem. The latter Time Dependent Processing Times,” Discrete
is considered as a particular case of Rm| |Cmax family Applied Mathematics, vol. 70, no. 1, pp. 81-93,
problems known as strongly NP-hard, therefore, a 1996. https://doi.org/10.1016/0166-
polynomial reduction should realize a significant 218X(96)00102-3
efficiency to treat these problems. As we will show, [9] Cheng T., Ding Q., and Lin B., “A Concise
BSUM is omnipresent in several kind of applications as Survey of Scheduling with Time-Dependent
manufacturing, transportation, logistic and routing. it is Processing Times,” Discrete Applied
of major importance in many company activities. The Mathematics, vol. 152, no. 1, pp. 1-13, 2004.
problem complexity and the optimality of the algorithm https://doi.org/10.1016/0166-218X(96)00102-3
are reported, proven and discussed. [10] Ebenlendr T., Kral M., and Sgall J., “Graph
Balancing: A Special Case of Scheduling
References Unrelated Parallel Machines,” in Proceedings of
the nineteenth Annual ACM-SIAM Symposium on
[1] Adan J., Adan I., Akcay A., Van den Discrete Algorithms, San Francisco, pp. 483-490,
Dobbelsteen R., and Stokkermans J., “A Hybrid 2008. DOI:10.1145/1347082.1347135
Genetic Algorithm for Parallel Machine [11] Fanjul-Peyro L. and Ruiz R., “Iterated Greedy
Scheduling At Semiconductor Back-End Local Search Methods for Unrelated Parallel
Production,” in Proceedings of the International Machine Scheduling,” European Journal of
Conference on Automated Planning and Operational Research, vol. 207, no. 1, pp. 55-69,
Scheduling,, vol. 28, pp. 298-302, 2018. 2010. https://doi.org/10.1016/j.ejor.2010.03.030
DOI: https://doi.org/10.1609/icaps.v28i1.13913 [12] Fanjul-Peyro L. and Ruiz R., “Scheduling
Exact Algorithm for Batch Scheduling on Unrelated Machine 623

Unrelated Parallel Machines with Optional Computing Environment,” The International

Machines and Jobs Selection,” Computers and Arab Journal of Information Technology, vol. 12,
Operations Research, vol. 39, no. 7, pp. 1745- no. 3, pp. 270-277, 2015.
1753, 2012. https://www.iajit.org/PDF/vol.12%2Cno.3/6131.
https://doi.org/10.1016/j.cor.2011.10.012 pdf
[13] Fanjul-Peyro L. and Ruiz R., “Size-Reduction [25] Ouazene Y. and Yalaoui F., “Identical Parallel
Heuristics for the Unrelated Parallel Machines Machine Scheduling with Time-Dependent
Scheduling Problem,” Computers and Processing Time,” Theoretical Computer
Operations Research, vol. 38, no. 1, pp. 301-309, Science, vol. 721, pp. 70-77, 2018.
2011. https://doi.org/10.1016/j.cor.2010.05.005 https://doi.org/10.1016/j.tcs.2017.12.001
[14] Gawiejnowicz S., Time-Dependent Scheduling, [26] Pinedo M., Scheduling, Theory, Algorithms, and
Springer, 2008. Systems, Springer Science+Business Media,
[15] Goldwasser M. and Pedigo M., “Online, Non- 2012. DOI 10.1007/978-3-319-26580-3
Preemptive Scheduling of Equal-Length Jobs on [27] Shchepin E. and Vakhania N., “An Optimal
Two Identical Machines,” in Proceedings of the Rounding Gives A Better Approximation for
10th Scandinavian Workshop on Algorithm Scheduling Unrelated Machines,” Operations
Theory (SWAT), Riga, pp. 113-123, 2006. Research Letters, vol. 33, no. 2, pp. 127-133,
[16] Ji M. and Cheng T., “Parallel-Machine 2005. https://doi.org/10.1016/j.orl.2004.05.004
Scheduling of Simple Linear Deteriorating Jobs,” [28] Vakhania N., Werner F., and Alberto J.,
Theoretical Computer Science, vol. 410, no. 38, “Scheduling Unrelated Machines with Two
pp. 3761-3768, 2009. Types of Jobs,” International Journal of
https://doi.org/10.1016/j.tcs.2009.04.018 Production Research, vol. 52, no. 13, pp. 3793-
[17] Kononov A. and Gawiejnowicz S., “Np-Hard 3801, 2014.
Cases in Scheduling Deteriorating Jobs on DOI:10.1080/00207543.2014.888789
Dedicated Machines,” The Journal of the [29] Yin N., Kang L., Sun T., Yue C., and Wang R.,
Operational Research Society, vol. 52, no. 6, pp. “Unrelated Parallel Machines Scheduling with
708-717, 2001. Deteriorating Jobs and Resource Dependent
https://www.jstor.org/stable/254283 Processing Times” Applied Mathematical
[18] Kravchenko S. and Werner F., “Parallel Machine Modelling, vol. 38, no. 19, pp. 4747-4755, 2014.
Problems with Equal Processing Times: A https://doi.org/10.1016/j.apm.2014.03.022
Survey,” Journal of Scheduling, vol. 14, pp. 435-
444, 2011. DOI:10.1007/s10951-011-0231-3 Hemmak Allaoua associate
[19] Lenstra J., Shmoys D., and Tardos E., professor at department of computer
“Approximation Algorithms for Scheduling science, faculty of mathematics and
Unrelated Parallel Machines,” Mathematical informatics since, Mohamed
Programming, vol. 46, pp. 259-271, 1990. Boudiaf University of M’sila,
[20] Li S. and Yuan J., “Parallel-Machine Scheduling Algeria since 2007. Having taught
with Deteriorating Jobs and Rejection,” several syllabuses such as
Theoretical Computer Science, vol. 411, no. 40, combinatorial optimization, metaheuristics, network,
pp. 3642-3650, 2010. language theory. Director of laboratory of informatics
https://doi.org/10.1016/j.tcs.2010.06.008 and its application of M’sila (LIAM) and head of the
[21] Mokotoff E., “Parallel Machine Scheduling team of optimization and artificial intelligence. Having
Problems: A Survey,” Asia-Pacific Journal of several publications, talks, activities in these research
Operational Research, vol. 18, no. 2, pp. 193- fields. Chair of ISIA’20 (International Symposium of
202, 2001. Informatics and its Application). Actually, TPC
[22] Moser M., Musliu N., Schaerf A., and Winter F., member of several conferences and reviewer for
“Exact and Metaheuristic Approaches for several journals in the field.
Unrelated Parallel Machine Scheduling,” Journal
of Scheduling, vol. 25, no. 5, pp. 507-534, 2022.
https://doi.org/10.1007/s10951-021-00714-6
[23] Mosheiov G., “Multi-Machine Scheduling with
Linear Deterioration,” Information Systems and
Operational Research, vol. 36, no. 4, pp. 205-
214, 1998.
https://doi.org/10.1080/03155986.1998.11732359
[24] Munir E., Ijaz S., Anjum S., Khan A., Anwar W.,
and Nisar W., “Novel Approaches for Scheduling
Task Graphs in Heterogeneous Distributed