• 전기의세계
• /
• v.27 no.4
• /
• pp.41-44
• /
• 1978

Intermittent failures exercise their effects only part of the time but constitute a dominant factor for the field failures. We consider the data raliability of the partially self-checking network with which a single intermittent failure will be recovered by a rollback method. Even if the self-testingness of partially self-checking network is guranteed for a set of permanent failures, it sometimes may not be so for intermittent failures. We introduce the notion of error residual and provide the basis for calculating the data reliability. Both the duration of each intermittent failure and the occurrence interval of successive ones are assumed to be negative exponentially distributed; the convolution of the intervals is distributed according to an Erlangen distribution.

• Transactions of Materials Processing
• /
• v.8 no.1
• /
• pp.92-100
• /
• 1999

A procedure is proposed for determining the initial thickness distribution in oder to produce a specified final thickness distribution for the axisymmetrical superplastic blow forming processes. Weighted parameter is introduced to improve the simple ad $d_traction method and the initial blank thickness distribution is obtained by optimizing the weighted parameter. This method is applied to superplastic free bulging process with the uniform thickness distribution of final shape to confirm its validity. The optimum initial blank thickness distributions is obtained from arbitrary axisymmetrical superplastic blow forming processes such as dome, cone and cylindrical cup forming with die contact. It is concluded that the ad $d_traction method with weighted parameter is an effective method for an optimum blank thickness distribution design.esign.

• The Transactions of the Korea Information Processing Society
• /
• v.7 no.7
• /
• pp.2008-2019
• /
• 2000

In this paper, we propose a recovery method for high dimensional index structures. It recovers efficiently transactions including reinsert operations that needs undo or rollback due to system failures or transaction failures. It is based on WAL(Write Ahead Logging) protocol. We apply the method to the FCIR-Tree and implement it based on MiDAS-III which is the storage system of a multimedia DBMS, called BADA-III. We also show through performance evaluation that the recovery method with our algorithm recovers reinsert operations efficiently over that without our algorithm.

• ETRI Journal
• /
• v.46 no.1
• /
• pp.59-70
• /
• 2024

We introduce a high-performance named entity recognition (NER) model for written and spoken language. To overcome challenges related to labeled data scarcity and domain shifts, we use transfer learning to leverage our previously developed KorBERT as the base model. We also adopt a meta-pseudo-label method using a teacher/student framework with labeled and unlabeled data. Our model presents two modifications. First, the student model is updated with an average loss from both human- and pseudo-labeled data. Second, the influence of noisy pseudo-labeled data is mitigated by considering feedback scores and updating the teacher model only when below a threshold (0.0005). We achieve the target NER performance in the spoken language domain and improve that in the written language domain by proposing a straightforward rollback method that reverts to the best model based on scarce human-labeled data. Further improvement is achieved by adjusting the label vector weights in the named entity dictionary.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.57 no.5
• /
• pp.875-880
• /
• 2008

In this paper, a novel method to determine the optimal checkpoint interval for real-time control task is proposed considering its performance degradation according to tasks's execution time. The control task in this paper has a specific sampling period shorter than its deadline. Control performance is degraded as the control task execution time is prolonged across the sampling period and eventually zero when reached to the deadline. A new performance index is defined to represent the performance variation due to the extension of task execution time accompanying rollback fault recovery. The procedure to find the optimal checkpoint interval is addressed and several simulation examples are presented.

• Proceedings of the Korean Information Science Society Conference
• /
• 2000.04b
• /
• pp.128-130
• /
• 2000

B+-Tree는 데이트베이스 관리 시스템에서 대용량의 데이터를 관리하기 위해 가장 널리 사용되는 인덱스이다. 그런 기존의 B+-Tree는 데이터베이스의 초기 구성 및 재구성시 많은 비용이 들고, 또한 삭제 연산의 빈번한 발생시 색인 구조 변경연산의 발생빈도가 높아져 동시성이 떨어진다는 단점이 있다. 이러한 문제점을 해결하기 위해서 기존 대부분의 데이터베이스 관리시스템에서는 일괄구성과 지연삭제를 이용하고 있으나, 동시성 및 회복에 대한 처리가 미흡하여 실제 시스템에 적용하기에는 문제가 있다. 따라서 본 논문에서는 일괄구성과 지연삭제 방법을 적용한 B+-Tree에서의 동시성 및 회복기법을 제안한다. 제안된 기법은 일괄구성 시에 잠금의 부하와 연속적인 철회(Cascade Rollback)가 없고, 또한 지연 삭제기법을 확장함으로써 빈 페이지 리스트 관리에 대한 부하가 없으며, 삭제 연산에 대한 회복 시 논리적 복귀(Logical Undo)가 빨라지고 구현이 간단해진다는 장점이 있다.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.60 no.1
• /
• pp.193-200
• /
• 2011

This paper presents an optimal checkpoint strategy for fault-tolerance in real-time systems. In our environment, multiple real-time tasks with arbitrary periods are scheduled in the system by Rate Monotonic (RM) algorithm, and checkpoints are inserted at a constant interval in each task while the width of interval is different with respect to the task. We propose a method to determine the optimal checkpoint interval for each task so that the probability of completing all the tasks is maximized. Whenever a fault occurs to a checkpoint interval of a task, the execution time of the task would be prolonged by rollback and re-execution of checkpoints. Our scheme includes the schedulability test to examine whether a task can be completed with an extended execution time. A numerical experiment is conducted to demonstrate the applicability of the proposed scheme.

• The KIPS Transactions:PartC
• /
• v.16C no.1
• /
• pp.65-72
• /
• 2009

Due to the popularity of mobile devices and advances in wireless communication technologies, a mobile ad hoc network has received a lot of attention. In the existing data replication management research, the use of a replica has been shown to be an efficient technique for improving data accessibility. However, to use a replica in ad hoc networks, the data consistency between the original data and the replica should be guaranteed. In the traditional research, a mobile node should check an original data whether the data is updated or not. However, It may be costly or sometimes impossible to check the original data. In the case of the time constraint applications, the checking cost can cause more serious problem. In this paper, we propose the time-guarantee based cache invalidation schemes for time constraint applications and the threshold based compensation method to enhance the time-guarantee based scheme. The proposed schemes can remove the "rollback" problem. Simulation results show that our schemes outperform the previous ones in terms of access time with little loss of data currency.

• Journal of Institute of Control, Robotics and Systems
• /
• v.17 no.11
• /
• pp.1074-1081
• /
• 2011

As robotics systems are becoming more complex there is the need to promote component based robot development, where systems can be constructed as the composition and integration of reusable building block. One of the most important challenges facing component based robot development is safeguarding against software component failures and malfunctions. The health monitoring of the robot software is most fundamental factors not only to manage system at runtime but also to analysis information of software component in design phase of the robot application. And also as a lot of monitoring events are occurred during the execution of the robot software components, a simple data treatment and efficient memory management method is required. In this paper, we propose an efficient events monitoring and data management method by modeling robot software component and monitoring factors based on robot software framework. The monitoring factors, such as component execution runtime exception, Input/Output data, execution time, checkpoint-rollback are deduced and the detail monitoring events are defined. Furthermore, we define event record and monitor record pool suitable for robot software components and propose a efficient data management method. To verify the effectiveness and usefulness of the proposed approach, a monitoring module and user interface has been implemented using OPRoS robot software framework. The proposed monitoring module can be used as monitoring tool to analysis the software components in robot design phase and plugged into self-healing system to monitor the system health status at runtime in robot systems.

• Journal of KIISE:Information Networking
• /
• v.29 no.5
• /
• pp.562-572
• /
• 2002

Distributed object technologies have become popular in developing distributed systems. Although such middleware platforms as DSOM, DCOM, CORBA and Java RMI ease the development of distributed applications, they do not directly improve the reliability and the availability of these applications. Because the task of developing fault-tolerance techniques for distributed object paradigms is often complicated and error-prone, there is a great need for a development toolkit that enhances the reliability and the availability of distributed objects. In this paper, we propose a fault-tolerant distributed middleware system based on RMI, called IMMORTAL. We use a log-based rollback-recovery mechanism for supporting reliable distributed computing. Through a series of experiments, we observe that benchmark applications on the IMMORTAL tolerate hardware and software failures and evaluate its performance and scalability.