• 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 KIISE:Computer Systems and Theory
• /
• v.28 no.12
• /
• pp.660-673
• /
• 2001

In Video-on-Demand (VoD) servers, disk throughput is an important system design parameter because it is directly related to the number of user requests that can be served simultaneously. In this paper, we propose an efficient periodic request grouping scheme for disk array-based VoD servers that reduces the disk seek time, thus improving the disk throughput of VoD disk arrays. To reduce the disk seek time, the proposed scheme groups the periodic requests that access data blocks stored in adjacent regions into one, and arranges these groups in a pre-determined order (e.g., in left-symmetric or right-symmetric fashion). Our simulation result shows that the proposed scheme reduces the average disk bandwidth required by a single video stream and can serve more user requests than existing schemes. For a data block size of 192KB, the number of simultaneously served user requests is increased by 8% while the average waiting time for a user request is decreased by 20%. We also propose an adaptation technique that conforms the proposed scheme to the user preference changes for video streams.

• Journal of Digital Contents Society
• /
• v.14 no.4
• /
• pp.471-479
• /
• 2013

WiBro technology provide mobile devices with a transmission speed of more than 1 Mbps even under the condition of high mobility (120 km/h). In a building, however, there is more demand for high bandwidth for various activities, and WiBro services may not be available. In order to guaranteeing a seamless service continuity and high transmission capability for mobile users within a specific establishment that has deployed WiBro network, we should secure an internetworking platform for WiBro and WiFi networks. This paper introduces WiMax forum and 3GPP standards, proposes WiBro-WiFi Interworking (WWI) architecture, its implementation details, and finally verifies validity of the proposal by using OPNet simulations.

• The Journal of the Korea Contents Association
• /
• v.16 no.5
• /
• pp.251-258
• /
• 2016

Recently, the demand for fast storage devices is rapidly increasing in cloud platforms, social network services, etc. Despite the development of fast storage devices, the traditional Linux I/O stack is not able to exploit the full extent of the performance improvement since it has been optimized for disk-based storage devices. In this paper, we propose an optimized I/O stack which can fully utilize the I/O bandwidth and latency of fast storage devices. To this end, we design a new I/O interface to replace the current block I/O interface and optimize our I/O interface. Our optimized I/O interface bypasses operations/layers in block I/O subsystems of the current Linux I/O stack to fully exploit fast storage devices. We also optimize the Linux file systems such as ext2 and ext4 to run on our I/O interface. We evaluate our I/O stack with multiple benchmarks and the experimental results show that our I/O stack achieves 1.7 times better throughput compared to traditional Linux I/O stack.

• The KIPS Transactions:PartC
• /
• v.16C no.5
• /
• pp.605-614
• /
• 2009

Despite its graceful bandwidth saving nature, IP multicast has not been successful in widely spreading and using. Challenging includes policy and technological issues such as inter-domain routing, multicast availability and reachability. Properly detecting and isolating the faults would be the first step to stabilize IP multicast. In this paper, we introduce a scalable multicast monitoring framework. To efficiently cope with multicast network and networking problems, it enlarges its monitoring scopes from collecting delivery statistics to verifying end-to-end multicast availability, reachability and interactivity. We carry out various networking experiments to verify the scalability and feasibility of the proposed framework.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.45 no.6
• /
• pp.33-44
• /
• 2008

Multipath routing protocols with load balance, fault tolerance, aggregated bandwidth in Ad-hoc Networks provide improved throughput and reliable route as compared with singlepath routing protocols. However, multipath routing protocols have not been explored thoroughly in the domain of overhead in Ad-hoc Networks. In this paper, we analyze and compare on-demand singlepath and multipath routing with IEEE 802.11 DCF in terms of Routing overhead and MAC overhead. The results reveals that in comparison with singlepath routing protocol, multipath routing mechanism creates more overheads but provides better performance in congestion and capacity provided that the route length is within a certain upper bound which is derivable. The analytical results are further confirmed by simulation.

• Journal of KIISE:Information Networking
• /
• v.29 no.3
• /
• pp.314-323
• /
• 2002

The routing protocols designed for wired networks can hardly be used for mobile ad-hoc networks due to limited bandwidth of wireless transmission and unpredictable topological change. Recently, several routing protocols for mobile ad-hoc networks have been Proposed. However, when theme protocols are applied to support real time services like multimedia transmission, they still have problems in ad-hoc networks, where the topology changes drastically. In this paper, we propose a new route selection algorithm which selects the most reliable route that is impervious to route failures by topological changes by mobile hoots. For reliable route selection, the concept of virtual zone (stable zone and caution zone) is proposed. The zone is located in a mobile node'transmission range and determined by mobile node's mobility information received by Global Positioning System (GPS). The proposed algorithm is applied to the route discovery procedure of the existing on-demand routing protocol, AODV, and evaluated by simulation in various traffic conditions and mobility patterns.

• Journal of KIISE:Information Networking
• /
• v.29 no.3
• /
• pp.224-232
• /
• 2002

The routing protocols designed for wired networks can hardly be used for mobile ad-hoc networks due to limited bandwidth of wireless transmission and unpredictable topological change. Recently, several routing protocols for mobile ad-hoc networks have been proposed. However, when these protocols are applied to support real time services like multimedia transmission, they still have problems in ad-hoe networks, where the topology changes drastically. In this paper, we propose a new route selection algorithm which selects the most reliable rouse that is impervious to route failures by topological changes by mobile hosts. For reliable route selection, the concept of virtual zone (stable lone and caution zone) is proposed. The lone is located in a mobile node's transmission range and determined by mobile node's mobility information received by Global Positioning System (GPS). The proposed algorithm is applied to the route discovery procedure of the existing on-demand routing protocol, AODV, and evaluated by simulation in various traffic conditions and mobility patterns.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.51 no.12
• /
• pp.30-45
• /
• 2014

Recently, multimedia application users who demand for ubiquitous computing environment are rapidly increasing, and wireless mesh network is receiving attention as a cost-effective key technology for next generation wireless networking. When multiple flows are transmitting data at the same time in the network, routing for path selection of each flow and link resource allocation for data transmission of each flow are one of the key factors that influence to the effectiveness of the network directly. In this paper, we consider problems for path discovery and resource allocation of links at the same time and we propose an algorithm based on mathematical modeling using a technique for cross-layer optimization design in STDMA-based wireless mesh networks that can enhance transfer performance for each flow. We show by performance analysis that the proposed algorithm can enhance the throughput performance by maximally utilizing given bandwidth resources when the number of flows increase in multi-hop wireless mesh networks.

• Journal of KIISE:Information Networking
• /
• v.36 no.3
• /
• pp.198-203
• /
• 2009

With the growing demand for multimedia services and advances in display technology, new applications for 3$\sim$D scene communication have emerged. While multi-view video of these emerging applications may provide users with more realistic scene experience, drastic increase in the bandwidth is a major problem to solve. In this paper, we propose a fast prediction mode decision algorithm which can significantly reduce complexity and time consumption of the encoding process. This is based on the object segmentation, which can effectively identify the fast moving foreground object. As the foreground object with fast motion is more likely to be encoded in the view directional prediction mode, we can properly limit the motion compensated coding for a case in point. As a result, time savings of the proposed algorithm was up to average 45% without much loss in the quality of the image sequence.