HongSoo Kim’s research while affiliated with Samsung Techwin Co. and other places

A mobile agent system is regarded as an attractive technology when developing distributed applications. However, mobility makes it more difficult to trace agents. It is also more complex for agents to communicate with each other in a reliable manner. Therefore, a reliable communication protocol is necessary to control and monitor mobile agents and deliver messages between them. In this paper, a new reliable communication protocol (RCP) is proposed for a multiregion mobile agent computing environment. RCP is implemented on the ODDUGI mobile agent system. Analysis and evaluation show that RCP fulfills the following design goals: reliability, asynchrony, timeliness, location dependency, scalability, and communication cost.

In this paper we discuss the design of result verification in desktop grid systems. In this design, correctness and performance are considered as important issues. To guarantee the correctness of work results, sabotage-tolerant mechanisms have been mainly used, such as voting-based schemes and trust-based schemes. However, these mechanisms result in low scalability and high computation delay because they can not cope effectively with dynamic environments. In this paper, we propose a Sabotage-Tolerant Scheduling for Result Verification (STSRV), which is based on mobile agent technology. In STSRV, mobile agents are used to check periodically the credibility and availability of each volunteer. Using credibility and availability information, our desktop grid system can provide correctness of work results without a huge increase in the computation delay caused by result verification. Additionally, simulation results show that STSRV increases turnaround time for works from the viewpoint of credibility and availability, and thus enhances the overall performance of our desktop grid systems.

A mobile agent is regarded as an attractive technology when developing distributed applications in mobile and ubiquitous comput- ing environments. In this paper, we present ODDUGI, a java-based ubiquitous mobile agent system. The ODDUGI mobile agent system provides fault tolerance, security, location management and message delivery mechanisms in a multi-region mobile agent computing environ- ment. We describe the architecture, design concepts and main features of the ODDUGI. In addition, we present the One-Touch Campus Service application developed on the basis of ODDUGI in mobile and ubiquitous computing environments.

This chapter reviews dynamism in desktop Grid computing and explains the advanced stochastic scheduling scheme with the Markov Job Scheduler based on Availability (MJSA) in the environment. In recent years, Grid computing [1] has received considerable interest in the field of academics and enterprise. Numerous attempts have been made to organize cost efficient large-scale Grid computing. Desktop Grid computing [13,19,2] is a more flexible paradigm that is used to achieve high performance and high throughput with desktop resources that are less stable and has more inferior performance compared to traditional Grid. It is comprised of a diverse set of desktops interconnected with various network forms ranging from Local Area Network (LAN) to the Internet. Desktop Grid system has played a leading role in the development of large scale aggregated computing power harvested from the edge of the Internet at lower cost. The main goals of the system are to accomplish high throughput and performance by mobilizing the potential colossal computational resources of idle desktops. However, since a desktop peer is a fluctuating resource that connects to the system, performs computations and disconnects to the network at will, desktop volatility makes the system unstable and unreliable. To develop a reliable desktop Grid computing system, a scheduling scheme must consider the dynamic nature (i.e., volatility) of volunteers and a resource selection scheme should adapt to such a dynamic environment, as the selection is getting complicated due to the uncertain behavior of desktops. This chapter demonstrates desktop state change modelling and an advanced resource selection scheme, Selection of Credible Resource with Elastic Window (SCREW), to choose reliable resources in dynamic computational desktop Grid environments. Markov modelling of the dynamic state turning provides understanding of the pattern of desktop behavior while SCREW selects qualified desktops that satisfy time requirements to complete given workloads and adapts to the needs of the user and the application on the fly.

In this paper, we discuss the design of replication mechanism to guarantee correctness and support deadline tasks in desktop grid systems. Both correctness and performance are considered important issues in the design of such systems. To guarantee the correctness of results, voting-based and trust-based sabotage tolerance mechanisms are generally used. However, these mechanisms suffer from two potential shortcomings: waste of resources due to running redundant replicas of the task, and increase in turnaround time due to the inability to deal with dynamic and heterogeneous environments. In this paper, we propose a Fuzzy-based Adaptive Replication Mechanism (FARM) for sabotage-tolerance with deadline tasks. This is based on the fuzzy inference process according to their trusty and result-return probability. Using these two parameters, our desktop grid system can provide both the sabotage-tolerance and a reduction in turnaround time. In addition, simulation results show that compared to existing mechanisms, the FARM can reduce resource waste in replication without increasing the turnaround time.

Although desktop Grid computing has been regarded as a cost-efficient computing paradigm, the system has suffered from scalability issues caused by its centralized structure. In addition, resource volatility generates system instability and performance deterioration. However, regarding the provision of a reliable and stable execution environment, resource management becomes more intricate when the system is constructed in a fully decentralized fashion without a central server. Scaling the system numerically and geographically is necessary for autonomous network organization, facile adaptation to execution failure and dynamic self-management of volatile resources. In order to develop a fully decentralized desktop Grid computing system securely, we propose an autonomous desktop Grid computing system, Self-Gridron based on a neural overlay network. Self- Gridron supports reliable, autonomous, and cost-effective scheduling which includes eligible resource classification and job management (i.e. allocation, replication, and reassignment). Furthermore, Self-Gridron provides sovereign learning with error correction) and evolves adaptively by itself to system changes or failure on the fly while improving performance.

Desktop Grid has recently received the strong attraction for executing high throughput applications as CPU, storage and network capacities improve and become cheaper. Desktop Grid is different from Grid in many respects, but there is no general survey or taxonomy for desktop Grid. Therefore, we propose a new comprehensive taxonomy and survey of desktop Grid in order to characterize and categorize desktop Grid.

Agents' mobility makes it difficult for them to deliver messages reliably, but a new system could change that. To asynchronously deliver messages to a mobile agent, the reliable asynchronous message delivery (RAMD) protocol places a "blackboard" in each region server for sharing information. RAMD also relates message delivery with a mobile agent's migration, thus addressing the tracking and message-delivery problems'