• International Journal of Contents
• /
• v.7 no.2
• /
• pp.1-5
• /
• 2011

Allocation of computing resources is a crucial issue when dealing with a huge number of tasks to be completed according to a given deadline and cost constraints. The task scheduling to several resources (e.g. grid, cloud or a supercomputer) with different characteristics is not trivial, especially if a trade-off in terms of time and cost is considered. We propose an allocation approach able to fulfill the given requirements about time and cost through the use of optimizing techniques and an adaptive behavior. Simulated productions of tasks have been run in order to evaluate the characteristics of the proposed approach.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.20 no.1
• /
• pp.135-140
• /
• 2020

Recently, due to the rapid diffusion of intelligent systems and IoT technologies, power saving techniques in real-time embedded systems has become important. In this paper, we propose P-GA (Parallel Genetic Algorithm), a scheduling algorithm aims at reducing the power consumption of real-time systems in multi-core hybrid memory environments. P-GA improves the Proportional-Fairness (PF) algorithm devised for multi-core environments by combining the dynamic voltage/frequency scaling of the processor with the nonvolatile memory technologies. Specifically, P-GA applies genetic algorithms for optimizing the voltage and frequency modes of processors and the memory types, thereby minimizing the power consumptions of the task set. Simulation experiments show that the power consumption of P-GA is reduced by 2.85 times compared to the conventional schemes.

• Proceedings of the Korean Information Science Society Conference
• /
• 2007.06b
• /
• pp.464-469
• /
• 2007

현재 이동용 장치(Mobile Device)들에서 전력 소모는 사용자들의 요구에 따라 성능 다음으로 중요한 비중을 차지하고 있다. 특히 배터리 셀의 기술 증가에 비해 프로세서들의 성능 및 요구하는 소비전력이 크게 증가함에 따라 프로세서의 전력 소모를 최소화 하는 연구들이 많이 진행되고 있다. 특히 프로세서의 전력 소모가 많은 비중을 차지함에 따라 프로세서의 전력 소모를 낮추기 위한 방법으로 많은 프로세서들은 DVS(Dynamic Voltage Scaling)와 DFS(Dynamic Frequency Scaling)를 지원한다. 실제 프로세서의 전력 소모는 공급전압에 의 제곱에 비례하고 동작 클럭(Clock) 주파수에 비례한다. 그러나 공급전압은 다시 동작 클럭 주파수에 비례함으로써 DVS와 DFS를 지원하는 대부분의 프로세서는 동작 클럭 주파수를 낮춤으로서 많은 전력 소모를 줄일 수 있게 된다. 그러나 동작 클럭 주파수를 낮추게 되면 태스크들의 실행 시간이 길어지게 되어 실시간 시스템에서 실시간성을 보장하지 못하게 된다. 본 논문에서는 상호간에 공유자원을 갖는 태스크들의 실시간성을 보장하며 동작 클럭 주파수를 낮추는 알고리즘을 제안한다.

• The Journal of the Korea Contents Association
• /
• v.4 no.2
• /
• pp.50-60
• /
• 2004

Recently, most of the embedded system is required not only many functions but also real-time characteristics in purpose. In the hard real-time system, especially, strict deadline of periodic task can affect the performance of the system. In this paper, we design and implement the scheduler based on RM(Rate-Monotonic) rule. This scheduler makes feasible patterns based on EDF(Earliest deadline first) rule with extended schedulability inspection before execution, for periodic task-set that has high CPU utilization and then, execute periodic task-set depended on feasible patterns. The feasible pattern formed into EDF rule is capable of the efficiency of CPU up to 100 percentage and by the referenced execution of the feasible pattern is possible of removing the red-time scheduling overhead that is the defect of the order of dynamic assignment rule.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.32 no.8B
• /
• pp.509-520
• /
• 2007

In this paper, we propose a real-time sensor node platform that guarantees the real-time scheduling of periodic and aperiodic tasks through a multitask-based software decomposition technique. Since existing sensor networking operation systems available in literature are not capable of supporting the real-time scheduling of periodic and aperiodic tasks, the preemption of aperiodic task with high priority can block periodic tasks, and so periodic tasks are likely to miss their deadlines. This paper presents a comprehensive evaluation of how to structure periodic or aperiodic task decomposition in real-time sensor-networking platforms as regard to guaranteeing the deadlines of all the periodic tasks and aiming to providing aperiodic tasks with average good response time. A case study based on real system experiments is conducted to illustrate the application and efficiency of the multitask-based dynamic component execution environment in the sensor node equipped with a low-power 8-bit microcontroller, an IEEE802.15.4 compliant 2.4GHz RF transceiver, and several sensors. It shows that our periodic and aperiodic task decomposition technique yields efficient performance in terms of three significant, objective goals: deadline miss ratio of periodic tasks, average response time of aperiodic tasks, and processor utilization of periodic and aperiodic tasks.

• Journal of KIISE:Information Networking
• /
• v.33 no.2
• /
• pp.113-130
• /
• 2006

Recently with the explosive growth of Internet applications, the attacks of hackers on network are increasing rapidly and becoming more seriously. Thus information security is emerging as a critical factor in designing a network system and much attention is paid to Network Intrusion Detection System (NIDS), which detects hackers' attacks on network and handles them properly However, the performance of current intrusion detection system cannot catch the increasing rate of the Internet speed because most of the NIDSs are implemented by software. In this paper, we propose a new high performance network intrusion using Network Processor. To achieve fast packet processing and dynamic adaptation of intrusion patterns that are continuously added, a new high performance network intrusion detection system using Intel's network processor, IXP1200, is proposed. Unlike traditional intrusion detection engines, which have been implemented by either software or hardware so far, we design an optimized architecture and algorithms, exploiting the features of network processor. In addition, for more efficient detection engine scheduling, we proposed task allocation methods on multi-processing processors. Through implementation and performance evaluation, we show the proprieties of the proposed approach.

• 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%.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.7
• /
• pp.2977-2997
• /
• 2018

In big data era, fresh data grows rapidly every day. More than 30,000 gigabytes of data are created every second and the rate is accelerating. Many organizations rely heavily on real time streaming, while big data stream computing helps them spot opportunities and risks from real time big data. Storm, one of the most common online stream computing platforms, has been used for big data stream computing, with response time ranging from milliseconds to sub-seconds. The performance of Storm plays a crucial role in different application scenarios, however, few studies were conducted to evaluate the performance of Storm. In this paper, we investigate the performance of Storm under different application scenarios. Our experimental results show that throughput and latency of Storm are greatly affected by the number of instances of each vertex in task topology, and the number of available resources in data center. The fault-tolerant mechanism of Storm works well in most big data stream computing environments. As a result, it is suggested that a dynamic topology, an elastic scheduling framework, and a memory based fault-tolerant mechanism are necessary for providing high throughput and low latency services on Storm platform.

• The Transactions of the Korea Information Processing Society
• /
• v.6 no.9
• /
• pp.2320-2331
• /
• 1999

Over the Internet, in order to utilize a collection of idle computers as a parallel computing platform, we propose a new scheme called GICE(Global Internet Computing Environment). GICE is motivated to obtain high programmability, efficient support for heterogeneous computing resources, system scalability, and finally high performance. The programming model of GICE is based on a single address space. GICE is featured with a Java based programming environment, a dynamic resource management scheme, and efficient parallel task scheduling and execution mechanisms. Based on a prototype implementation of GICE, we address the concept, feasibility, complexity and performance of Internet computing.

• Journal of KIISE:Computing Practices and Letters
• /
• v.15 no.4
• /
• pp.304-308
• /
• 2009

This paper presents the deadline driven CPU-Power management scheme for the Real-Time Embedded OS: named TMO-eCos. It used the scheduling scenarios generated by a task serialization technique for hard real- time TMO system. The serializer does a off-line analysis at design time with period, deadline and WCET of periodic tasks. Finally, TMO-eCos kernel controls the CPU speed to save the power consumption under the condition that periodic tasks do not violate deadlines. As a result, the system shows a reasonable amount of power saving. This paper presents all of these processes and test results.