• Journal of KIISE:Computer Systems and Theory
• /
• v.30 no.7_8
• /
• pp.352-358
• /
• 2003

A cluster of servers is the most promising solution to provide highly scalable and highly available network services with low implementation cost. One of popular solutions is using dispatcher as a request distributor. However this solution has the problem that is can be the bottleneck or single-point of failure point of itself. ONE-IP, the previous work, solves this serious problem by using broadcast message for distributing the request packet in LAN. Using simple scheduling technique, in addition, the overhead due to dispatching the request becomes minimal. However, as ONE-IP using very simple static scheduling algorithm, it would spread the request among servers unevenly and downgrades the cluster performance consequently. In this paper, we propose an improved TCP connection scheduling technique solving this unbalanced distributing problem. Using Round-Robin(RR) scheduling instead of hashing clients IP address, proposed technique can more evenly distribute incoming service requests among the servers. Experiments show average 3.8% improved performance in throughput compared to original ONE-IP technique.

• KIPS Transactions on Computer and Communication Systems
• /
• v.10 no.12
• /
• pp.311-318
• /
• 2021

With the rapid development of the Internet of Things(IoT) technology recently, multi-access edge computing(MEC) is emerged as a next-generation technology for real-time and high-performance services. High mobility of users between MECs with limited service areas is considered one of the issues in the MEC environment. In this paper, we consider a vehicle edge computing(VEC) environment which has a high mobility, and propose a task migration algorithm to decide whether or not to migrate and where to migrate using DQN, as a reinforcement learning method. The objective of the proposed algorithm is to improve the system throughput while satisfying QoS(Quality of Service) requirements by minimizing the difference between queueing delays in vehicle edge computing servers(VECSs). The results show that compared to other algorithms, the proposed algorithm achieves approximately 14-49% better QoS satisfaction and approximately 14-38% lower service blocking rate.

• The Journal of Korean Institute of Next Generation Computing
• /
• v.15 no.4
• /
• pp.18-29
• /
• 2019

A dynamic adaptive streaming server consumes high processor power because it handles a large amount of transcoding operations at a time. For this purpose, multi-processor architecture is mandatory for which effective transcoding task distribution strategies are essential. In this paper, we present the design and implementation details of the transcoding workload distribution schemes at a 2-tier (frontend node and backend node) transcoding server. For this, we implemented four schemes: 1) allocation of transcoding tasks to appropriate back-end nodes, 2) task scheduling in the back-end node and 3) the communication between front-end and back-end nodes. Experiments were conducted to compare the estimated and the actual power consumption in a real testbed to verify the efficacy of the system. It also proved that the system can reduce the load on each node to optimize the power and time used for transcoding.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.05a
• /
• pp.158-160
• /
• 2022

Many office workers spend a lot of time performing repetitive office tasks in inefficient ways. We developed a user-friendly chatbot system based on Kakaotalk to automate repetitive tasks and used the chatbot in the real workplace. In the process of operating the chatbot, if several people use the chatbot at the same time, the server was down or could not respond. To address these issues, we performed code migration of programs used by chatbot back-end servers and tried several ways to improve server performance such as database redesign and load balancing. To determine how much each method affects performance improvement, we measured total request per second and average latency. After that, we proposed ways to improve the problems of using the chatbot in the work environment.

• The KIPS Transactions:PartA
• /
• v.14A no.1 s.105
• /
• pp.63-72
• /
• 2007

This paper introduces a load-balancing algorithm focusing on distributing web requests evenly into the web cluster servers. The load-balancing algorithms based on conventional periodic load-information update mechanism are not suitable for dynamic page applications, which are common in Cyber Foraging services, due to the problems caused by periodic synchronized load-information updating and the difficulties of work load estimation caused by embedded executing scripts of dynamic pages. Update-on-Finish algorithm solves this problem by using non-periodic load-update mechanism, and the web switch knows the servers' real load information only after their reporting and then distributes new loads according to the new load-information table, however it results in much communication overhead. Our proposed mechanism improve update-on-finish algorithm by using K-Percents-Finish mechanism and thus largely reduce the communication overhead. Furthermore, we consider the different capabilities of servers with a threshold Ti value and propose a load-balancing algorithm for servers with various capabilities. Simulation results show that the proposed K-Percents-Finish Reporting mechanism can at least reduce 50% communication overhead than update-on-finish approach while sustaining better load balancing performance than periodic mechanisms in related work.

• The KIPS Transactions:PartA
• /
• v.14A no.6
• /
• pp.371-382
• /
• 2007

Modern server systems are usually composed in the form of cluster systems in order to serve not only as many users but also as many kinds of applications as possible. The progression of the cluster system architecture leads in a middleware approach based on the Java framework. The middleware approach alleviates the efforts for the construction and the management of a server system but still preserves its performance and applications on the server. In this research, we introduce a new clustering scheme for the easy construction and maintenance of a cluster server system with the Java Management Extensions. We first demonstrate the construction and configuration process. Our experiment sets can verify that it is easy to construct, expand and manage a middleware based cluster system as well as the applications which reside on it. In addition, we can achieve reasonable performance on our service oriented clustered system with the help of state-of-the-art middleware. The experimental results of performance demonstration contain the availability of a server, and the effectiveness of load balancing and scheduling mechanisms. Especially, our service oriented scheduling mechanism was shown to successfully manage load imbalance under the normal load and cope with the overloaded situations, compared with other known scheduling mechanisms.