• KIPS Transactions on Computer and Communication Systems
• /
• v.11 no.7
• /
• pp.205-216
• /
• 2022

One of the many tools used for distributed deep learning training is Kubeflow, which runs on Kubernetes, a container orchestration tool. TensorFlow jobs can be managed using the existing operator provided by Kubeflow. However, when considering the distributed deep learning training jobs based on the parameter server architecture, the scheduling policy used by the existing operator does not consider the task affinity of the distributed training job and does not provide the ability to dynamically allocate or release resources. This can lead to long job completion time and low resource utilization rate. Therefore, in this paper we proposes a new operator that efficiently schedules distributed deep learning training jobs to minimize the job completion time and increase resource utilization rate. We implemented the new operator by modifying the existing operator and conducted experiments to evaluate its performance. The experiment results showed that our scheduling policy improved the average job completion time reduction rate of up to 84% and average CPU utilization increase rate of up to 92%.

• Journal of Korea Trade
• /
• v.23 no.3
• /
• pp.52-67
• /
• 2019

Purpose - This study analyzes the effect of the authorized economic operator-mutual recognition arrangement (AEO-MRA) on the performance of Korean exporters and importers. The effect of the import-export companies' characteristics, such as annual sales, the number of foreign markets, and overseas experience, on the AEO-MRA is deduced; the relationship between this effect and firm performance is analyzed. Design/methodology - An empirical research model was constructed and analyzed using structural equation modeling. The effect of AEO-MRA on logistics and operational performance was derived from the aforementioned characteristics as leading factors of the AEO-MRA. The regulatory influence of cooperation with logistics companies was analyzed in the AEO-MRA effect on logistics performance. Thus, 172 valid samples were obtained from import-export companies certified by the AEO-MRA. Findings - Among the aforementioned characteristics, only "annual sales" has a positive effect on the AEO-MRA, whose effect enhances logistics and operational performances. The AEO-MRA effect did not directly affect operational performance. Owing to the adjustment effect analysis, the AEO-MRA effect and logistics performance relationship is strengthened if the cooperative relationship with the logistics company is higher than a certain level. If this cooperation falls below a certain level, the AEO-MRA effect on logistics performance reduces. Thus, logistics cooperation is an important factor in the AEO-MRA effect and logistics performance relationship. Originality/value - Hinging on the resource-based theory and relational viewpoint, an empirical model that explains the relationship between the AEO-MRA effect and firm performance is established.