• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.18 no.7
• /
• pp.1749-1773
• /
• 2024

Recent advances in deep neural networks (DNNs) have greatly improved the accuracy and universality of various intelligent applications, at the expense of increasing model size and computational demand. Since the resources of end devices are often too limited to deploy a complete DNN model, offloading DNN inference tasks to cloud servers is a common approach to meet this gap. However, due to the limited bandwidth of WAN and the long distance between end devices and cloud servers, this approach may lead to significant data transmission latency. Therefore, device-edge collaborative inference has emerged as a promising paradigm to accelerate the execution of DNN inference tasks where DNN models are partitioned to be sequentially executed in both end devices and edge servers. Nevertheless, collaborative inference in heterogeneous edge environments with multiple edge servers, end devices and DNN tasks has been overlooked in previous research. To fill this gap, we investigate the optimization problem of collaborative inference in a heterogeneous system and propose a scheme CIS, i.e., collaborative inference scheme, which jointly combines DNN partition, task offloading and scheduling to reduce the average weighted inference latency. CIS decomposes the problem into three parts to achieve the optimal average weighted inference latency. In addition, we build a prototype that implements CIS and conducts extensive experiments to demonstrate the scheme's effectiveness and efficiency. Experiments show that CIS reduces 29% to 71% on the average weighted inference latency compared to the other four existing schemes.

• Journal of Internet of Things and Convergence
• /
• v.10 no.4
• /
• pp.109-115
• /
• 2024

This paper proposes an interference recognition-based resource allocation (IaRA) scheme for very high frequency (VHF) data exchange-terrestrial (VDE-TER) to ensure the reliability of automatic identification system (AIS) message transmission in a coastal environment with dense vessels. To this end, IaRA conducts channel selection and slot allocation based on scheduling information obtained through monitoring VHF data exchange system (VDES) channels. The simulation results show that IaRA achieved an 8.79% higher packet delivery ratio (PDR) and a 9.69% lower channel interference ratio (CIR) compared to the VDE-TER resource allocation scheme of the existing VDES standard.

• ETRI Journal
• /
• v.24 no.4
• /
• pp.311-322
• /
• 2002

Presently, optical burst switching (OBS) technology is under study as a promising solution for the backbone of the optical Internet in the near future because OBS eliminates the optical buffer problem at the switching node with the help of no optical/electro/optical conversion and guarantees class of service without any buffering. To implement the OBS network, there are a lot of challenging issues to be solved. The edge router, burst offset time management, and burst assembly mechanism are critical issues. In addition, the core router needs data burst and control header packet scheduling, a protection and restoration mechanism, and a contention resolution scheme. In this paper, we focus on the burst assembly mechanism. We present a novel data burst generation algorithm that uses hysteresis characteristics in the queueing model for the ingress edge node in optical burst switching networks. Simulation with Poisson and self-similar traffic models shows that this algorithm adaptively changes the data burst size according to the offered load and offers high average data burst utilization with a lower timer operation. It also reduces the possibility of a continuous blocking problem in the bandwidth reservation request, limits the maximum queueing delay, and minimizes the required burst size by lifting up data burst utilization for bursty input IP traffic.

• Journal of Korea Multimedia Society
• /
• v.18 no.9
• /
• pp.1098-1104
• /
• 2015

With the widespread use of smartphones and IoT (Internet of Things), data are being generated on a large scale, and there is increased for the analysis of such data. Hence, distributed processing systems have gained much attention. Hadoop, which is a distributed processing system, saves the metadata of stored files in name nodes; in this case, the main problems are as follows: the memory becomes insufficient; load occurs because of massive small files; scheduling and file processing time increases because of the increased number of small files. In this paper, we propose a solution to address the increase in processing time because of massive small files, and thus improve the processing performance, using the Reuse JVM method provided by Hadoop. Through environment setting, the Reuse JVM method modifies the JVM produced conventionally for every task, so that multiple tasks are reused sequentially in one JVM. As a final outcome, the Reuse JVM method showed the best processing performance when used together with CombineFileInputFormat.

• Journal of Korea Multimedia Society
• /
• v.7 no.4
• /
• pp.496-504
• /
• 2004

This paper preposes a various policies and experiment results to select streams in a presentation to be transcoded for mobileterminals after distinguishing the characteristics of streams in the presentation. This applies to the presentation composed of various data streams on a PC server. The rationale of the research is that the stream selection polices for transcoding enhances the speed of presentation in mobile terminals considerably faster than in the case when every stream has to be transcoded. Thus, the thesis suggests a stream selection polices for transcoding based on EPOB (End Point of Over Bandwidth) and aims to lower the requiring bandwidth of the presentation to the network bandwidth as well as to minimize early delay time for the presentation to be played in mobile terminals.

• International journal of advanced smart convergence
• /
• v.5 no.1
• /
• pp.34-46
• /
• 2016

Networked Control System (NCS) has evolved in the past decade through the advances in communication technology. The problems involved in NCS are broadly classified into two categories namely network issues due to network and control performance due to system network. The network problems are related to bandwidth allocation, scheduling and network security, and the control problems deal with stability analysis and delay compensation. Various delays with variable length occur due to sharing a common network medium. Though most delays are very less and mostly neglected, the network induced delay is significant. It occurs when sensors, actuators, and controllers exchange data packet across the communication network. Networked induced delay arises from sensor to controller and controller to actuator. This paper presents an adaptive delay compensation process for efficient control. Though Smith predictor has been commonly used as dead time compensators, it is not adaptive to match with the stochastic behavior of network characteristics. Time delay adaptive compensation gives an effective control to solve dead time, and creates a virtual environment using the plant model and computed delay which is used to compensate the effect of delay. This approach is simulated using TrueTime simulator that is a Matlab Simulink based simulator facilitates co-simulation of controller task execution in real-time kernels, network transmissions and continuous plant dynamics for NCS. The simulation result is analyzed, and it is confirmed that this control provides good performance.

• Journal of Korea Multimedia Society
• /
• v.23 no.2
• /
• pp.247-254
• /
• 2020

The recent Internet environment is characterized by the explosion of certain types of data, as the data that people want is affected by certain issues. In this paper, we propose a load balancing technique that considers the data generation forms. The concept of this technique is to prioritize some type of data when it suddenly explodes. This is a technique to build an add-on middle box on a switch to monitor packets and give priority to a queue for load balancing. This technique worked when certain types of data exploded. SDN(Software Defined Networking) has the advantage of efficiently managing a number of network equipment. However, load balancing in the SDN environment has not been studied much. Applying the proposed load balancing technique in the SDN environment can save time and budget and easily implement our policies. When the proposed load balancing technique is applied to the SDN environment, it has been found that the techniques we want can be easily applied to the network systems, and that efficient data processing is possible when certain types of data explosion.

• Journal of Digital Contents Society
• /
• v.7 no.3
• /
• pp.207-210
• /
• 2006

A synchronized multimedia presentation consists of a collection of objects, with each object having one or more rendering intervals within the presentation timeline. These intervals specify the objects' start times and end times relative to the presentation timeline. In this paper we consider the problem of streaming a multimedia presentation from a server to a client over a bandwith-limited communication network. We suppose that each of the static objects is layered-encoded. For a given maximum delay, we consider the problem of finding the optimal number of quality of the presentation. We devise efficient algorithms for determining an optimal policy for several criteria. We also consider the problem of gradual rendering of objects after their start times.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2018.07a
• /
• pp.201-202
• /
• 2018

자동차 부품산업은 자동차산업의 후방산업으로 완성차 메이커 제조산업의 조립용인 중간재 산업으로 부품의 품질에 따라 완성차의 성능이 좌우된다. 자동차 산업구조상 완성차에 종속적인 사업구조로 부품사의 독자 성장이 어렵고 수익성이 완성차업체에 종속 된다. 자동차부품사의 대형화 및 생산공정 자동화 변화로 기존의 수직계열화된 부품 공급 관계에도 변화가 예상되며, 완성차업체는 통합적인 시스템 부품을 생산할 수 있는 글로벌 대형 부품업체에 의존하는 상황이 전개될 수도 있다. 이와 같은 위기와 변화를 극복하기 위해 본 연구에서는 산업구조분석을 자동차 부품산업에 적용하여 자동차부품산업의 경쟁력제고 전략을 도출하고자 한다.

• Journal of the Korea Safety Management & Science
• /
• v.4 no.1
• /
• pp.105-114
• /
• 2002

Nowadays, most companies seek after the greatest profit by means of offering the goods which consumers want timely and efficiently and minimizing the cost of inventory and distribution channel which amounts to the great portion of total cost. And according as EC(Electronic Commerce) market has increased rapidly, SCM(Supply Chain Management) for EC become one of the most important facts for companies, therefore this paper suggest SCM scheme which EC, especially B to C, is added to the existing commerce system. This paper uses internet for information integration of distribution channel which is away from one another and applies TOC(Theory of Constraint)'s DBR(Drum-Buffer-Rope) Scheduling for synchronization through the whole supply chain. It is possible to synchronize the whole supply chain by means of making the speed of manufacturing and distribution to be controlled by consumer's order which is received in distribution center, and inventory and loss of sales opportunities are minimized by constant. Buffer Management. If inventories in distribution center is short, then it needs to search CCR(Capacity Constraint Resource) in supply chain and to control the speed of manufacturing and distribution according to the ability of CCR. This paper applies PT(Partial Transshipment) strategy for Delivery from distribution center to store or cyber consumer. the strategy this paper suggests chooses neighbour area from area which each distribution center takes charge, and then makes product ordered by cyber consumer which lives in the chosen area to be delivered according to inventory of distribution center.