• Journal of Korean Institute of Industrial Engineers
• /
• v.20 no.2
• /
• pp.31-38
• /
• 1994

We consider a single-machine scheduling problem dealing with the manufacture of components for subsequent assembly into end products. Each product requires both unique components and common components, and each production requires a setup. By making some assumptions on the data and the availability of the components for assembly, Baker provides on efficient dynamic programming algorithm for obtaining the optimal schedule. In this paper we do not impose any requirement on the data, and we solve the more complicated batching and sequencing problem. We suggest a simple heuristic method that is efficient and finds solutions that are optimal or close to the optimal solution.

• Journal of Korean Institute of Industrial Engineers
• /
• v.32 no.3
• /
• pp.163-171
• /
• 2006

This paper considers a scheduling problem to minimize the sum of the associated scheduling (production/delivery times) cost and the delivery cost for an integrated system of a single production machine and various transportation vehicles with transportation mode selection allowed. Each transportation mode is provided with a fixed number of vehicles at the associated delivery time and cost. The proposed problem is characterized as being NP-hard. Some solution properties are also characterized. Therewith, three heuristic algorithms (called SPT-based, LWF-based and WSPT-based heuristic) and a branch-and-bound algorithm are derived. In order to evaluate the effectiveness and efficiency of the proposed algorithms, computational experiments are made with some numerical instances.

• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.27 no.2
• /
• pp.24-28
• /
• 2004

This paper considers a scheduling problem concerned with an assembly system where two components are first treated In their own parallel machines and then pulled to be assembled into a final product at a single assembly machine. The objective measure is the mean completion time of jobs(a finite number of products). Through characterizing solution properties, we obtain the worst case error bounds of an arbitrary permutation and a SPT based heuristic.

• Journal of Korean Institute of Industrial Engineers
• /
• v.32 no.2
• /
• pp.98-103
• /
• 2006

In this research, the single machine capacitated lot-sizing and scheduling problem with sequence- dependent setup costs and setup times (CLSPSD) is considered. This problem is the extension of capacitated lot-sizing and scheduling problem (CLSP) with an additional assumption on sequence-dependent setup costs and setup times. The objective of the problem is minimizing the sum of production costs, inventory holding costs and setup costs satisfying customers' demands. It is known that the CLSPSD is NP-hard. In this paper, the MIP formulation is presented. To handle the problem more efficiently, a conceptual model is suggested, and one of the well-known meta-heuristics, the simulated annealing approach is applied. To illustrate the performance of this approach, various instances are tested and the results of this algorithm are compared with those of the CLPEX. Computational results show that this approach generates optimal or nearly optimal solutions.

• Journal of Electrical Engineering and information Science
• /
• v.1 no.1
• /
• pp.118-128
• /
• 1996

In safety critical hard real-time systems, a timing fault may yield catastrophic results. In order to eliminate the timing faults from the fast responsive real-time control systems, it is necessary to schedule a code based on high precision timing analysis. Further, the schedulability enhancement by having multiple processors is of wide spread interest. However, although an instruction level parallel processing is quite effective to improve the schedulability of such a system, none of the real-time applications employ instruction level parallel scheduling techniques because most of the real-time scheduling models have not been designed for fine-grain execution. In this paper, we present a timing constraint model specifying high precision timing constraints, and a practical approach for constructing static schedules for a VLIW execution model. The new model and analysis can guarantee timing accuracy to within a single machine clock cycle.

• Journal of Korean Institute of Industrial Engineers
• /
• v.38 no.1
• /
• pp.31-40
• /
• 2012

Distributed arrival time control (DATC) is a distributed feedback control algorithm for real-time scheduling problems in dynamic operational environment. Even though DATC has provided excellent performance for dynamic scheduling problems, it can be improved by considering the following considerations. First, the original DATC heavily depends on the quality of initial solution. In this paper, well-known dispatching rules are incorporated DATC algorithm to enhance its performance. Second, DATC improves its solution with adjusting virtual arrival times of jobs to be scheduled in proportion to the gap between completion time and due date iteratively. Since this approach assigns the same weight to all gaps generated with iterations, it fails to utilize significantly more the latest information (gap) than the previous ones. To overcome this issue we consider exponential smoothing which enable to assign different weight to different gaps. Using these two consideration This paper proposes A-DATC (Advanced-DATC). We demonstrate the effectiveness of the proposed scheduling algorithm through computational results.

• ETRI Journal
• /
• v.45 no.2
• /
• pp.318-328
• /
• 2023

Recently, embedded systems, such as mobile platforms, have multiple processing units that can operate in parallel, such as centralized processing units (CPUs) and neural processing units (NPUs). We can use deep-learning compilers to generate machine code optimized for these embedded systems from a deep neural network (DNN). However, the deep-learning compilers proposed so far generate codes that sequentially execute DNN operators on a single processing unit or parallel codes for graphic processing units (GPUs). In this study, we propose PartitionTuner, an operator scheduler for deep-learning compilers that supports multiple heterogeneous PUs including CPUs and NPUs. PartitionTuner can generate an operator-scheduling plan that uses all available PUs simultaneously to minimize overall DNN inference time. Operator scheduling is based on the analysis of DNN architecture and the performance profiles of individual and group operators measured on heterogeneous processing units. By the experiments for seven DNNs, PartitionTuner generates scheduling plans that perform 5.03% better than a static type-based operator-scheduling technique for SqueezeNet. In addition, PartitionTuner outperforms recent profiling-based operator-scheduling techniques for ResNet50, ResNet18, and SqueezeNet by 7.18%, 5.36%, and 2.73%, respectively.

• Journal of the Korean Operations Research and Management Science Society
• /
• v.23 no.1
• /
• pp.29-41
• /
• 1998

This paper addresses the problem of scheduling a set of jobs with a common due data on a single machine. The objective is to minimize the sum of the earliness and tardiness of jobs subject to $T_{max}{\le}{\Delta}\;for{\Delta}{\ge}0$. Properties for the $MAD/T_{max}$ problem are found and the problem is shown to be NP-complete in the ordinary sense. According to the range of Δ, the problem can be solved in polynomial time. Also, some special cases where an optimal schedule is found in polynomial time are discussed.

• Journal of Korean Institute of Industrial Engineers
• /
• v.19 no.4
• /
• pp.125-135
• /
• 1993

An Unrelated Parallel Processing with Weighted jobs and Setup times scheduling prolem is studied. We consider a parallel processing in which a group of processors(machines) perform a single operation on jobs of a number of different job types. The processing time of each job depends on both the job and the machine, and each job has a weight. In addition each machine requires significant setup time between processing jobs of different job types. The performance measure is to minimize total weighted flow time in order to meet the job importance and to minimize in-process inventory. We present a 0-1 Mixed Integer Programming model as an optimizing algorithm. We also present a simple heuristic algorithm. Computational results for the optimal and the heuristic algorithm are reported and the results show that the simple heuristic is quite effective and efficient.

• International journal of advanced smart convergence
• /
• v.13 no.3
• /
• pp.143-149
• /
• 2024

As the size of data and models in machine learning training continues to grow, training on a single server is becoming increasingly challenging. Consequently, the importance of distributed machine learning, which distributes computational loads across multiple machines, is becoming more prominent. However, several unresolved issues remain regarding the performance enhancement of distributed machine learning, including communication overhead, inter-node synchronization challenges, data imbalance and bias, as well as resource management and scheduling. In this paper, we propose ParamHub, which utilizes orchestration to accelerate training speed. This system monitors the performance of each node after the first iteration and reallocates resources to slow nodes, thereby speeding up the training process. This approach ensures that resources are appropriately allocated to nodes in need, maximizing the overall efficiency of resource utilization and enabling all nodes to perform tasks uniformly, resulting in a faster training speed overall. Furthermore, this method enhances the system's scalability and flexibility, allowing for effective application in clusters of various sizes.