Kuo-Chih Chu

Dynamical combination of byte level and Sub-Packet level FEC in HARQ mechanism to reduce error recovery overhead on video streaming over wireless networks

Byte level Forward Error Correction (B-FEC) is efficient for recovery from uniform bit errors, but not suitable to handle recovery from burst bit errors. Conversely, Sub-Packet level Forward Error Correction (SP-FEC) can alleviate the problem of large encoding/decoding delay jitter in Packet level Forward Error Correction (P-FEC) to efficiently handle recovery from burst bit errors, but has large error recovery overhead as P-FEC for recovery from uniform bit errors. This paper proposes a dynamic combination of byte level and Sub-Packet level Forward Error Correction (BSP-FEC) in the Hybrid Automatic Repeat reQuest (HARQ) mechanism to reduce the error recovery overhead. BSP-FEC not only can solve the problems appearing in B-FEC and SP-FEC, but also can get the advantages of B-FEC and SP-FEC in the HARQ mechanism. BSP-FEC replaces the SP-FEC checksum with B-FEC and uses Automatic Repeat reQuest (ARQ) when the network condition deteriorates. BSP-FEC not only utilizes an overhead cost model to dynamically decide the SP-FEC parameter and the B-FEC bit rate according to network conditions, but also utilizes a time constraint model to decide the ARQ retry limit. BSP-FEC dynamically adjusts the FEC redundancy to save bandwidth and improves the Decodable Frame Rate (DFR) and the Peak Signal to Noise Ratio (PSNR) of the delivered video streaming. Accordingly, BSP-FEC can improve multimedia communication performance to both avoid network congestion and shorten end-to-end delay by decreasing effective packet loss rate and packet recovery overhead. Because of the low packet recovery overhead, furthermore, BSP-FEC allows applications to transmit more application data in networks with limited bandwidth. Considering the compatibility, BSP-FEC is implemented in the application layer as the end-to-end protection method to protect packets from errors in wired/wireless networks. Numerical and simulation experimental results show that BSP-FEC obtains better recovery efficiency with the minimum error recovery overhead.

Adaptive Combiner for MapReduce on cloud computing

MapReduce is a programming model to process a massive amount of data on cloud computing. MapReduce processes data in two phases and needs to transfer intermediate data among computers between phases. MapReduce allows programmers to aggregate intermediate data with a function named combiner before transferring it. By leaving programmers the choice of using a combiner, MapReduce has a risk of performance degradation because aggregating intermediate data benefits some applications but harms others. Now, MapReduce can work with our proposal named the Adaptive Combiner for MapReduce (ACMR) to automatically, smartly, and trainer for getting a better performance without any interference of programmers. In experiments on seven applications, MapReduce can utilize ACMR to get the performance comparable to the system that is optimal for an application.

Networking without Dynamic Host Configuration Protocol server in Ethernet and Wireless Local Area Network

A Dynamic Host Configuration Protocol (DHCP) server is a well-known server deployed at a network to manage Internet Protocol (IP) addresses temporarily rentable to hosts in the network. Besides, a DHCP server provides hosts with important network information such as the subnet mask and the gateway IP address. However, a DHCP server has many drawbacks and should not be considered necessary in each network. If no DHCP server exists in a network to serve hosts and no network administrator helps users manually configure the hosts, currently no practical solution can make the hosts access networks. In this paper, Automatic Host Configuration Mechanism (AHCM) is proposed to make a host access Ethernet and Wireless Local Area Network (WLAN) without a DHCP server. AHCM can automatically locate IP addresses usable to hosts and find network information in a transparent way without any user interference. Working like protocol software inside an Operating System (OS), AHCM has high compatibility because of neither modifying any application nor deploying a third party server in a network. Most importantly, AHCM does not have drawbacks in a DHCP server. AHCM is implemented in the protocol stack on Windows XP and tested in several experiments to identify its overheads and performances.

Client-side session splice approach, a novel approach to achieving seamless handoffs for multimedia applications in mobile computing

Thanks to the invention of mobile computing technology, people nowadays can have various entertainment experiences with multi-function devices and heterogeneous network interfaces. When using multi-function devices with heterogeneous network interfaces to download and play streaming data, they maybe want to switch the communication from a network interface to another one in a device (i.e., the intra-terminal handoff), or to the network interface in another device (i.e., the inter-terminal handoff). In the handoffs, they have to manually reset or initiate the application, and then re-subscribe to the streaming service because the communication bound to the old network interface is lost. They are interrupted for a long time when watching the show. They have to remember the point or the scene played in the application before the handoff in order to re-subscribe to the streaming service after the handoff. They suffer from the troubling manual operations, especially when the handoff happens frequently to the network interfaces having limited communication ranges. When roaming through networks, accordingly, they need a way to download and play the streaming data without interruption due to handoffs. They can use the proposed Client-side Session Splice Approach (CSSA) to achieve seamless handoffs for multimedia applications in mobile computing. They can rely on the CSSA to automatically finish the handoff between network domains or homogeneous communication media in a network interface, the handoff between network interfaces in a device, and the handoff between network interfaces in different devices. For achieving seamless handoffs, they can count on the CSSA to automatically re-subscribe the streaming service on behalf of the application and smartly download the streaming data at the point or the scene played in the application just before the handoff happens. They don’t need to worry about the compatibility of the CSSA because the CSSA keeps intact applications, network infrastructures, and streaming servers. They can understand the principle, the practicability, the functionality, and the overheads of the CSSA through this paper.

Smart MapReduce cloud: Applying extra processing to intermediate data on demand

Cloud computing is the emerging and attractive technology and provides users with various services in a pay-as-you-go manner. Cloud computing nowadays does not limit resources of the services in a cloud to the computers that are far away from users and connected to each other in a data center with high speed networks at the same geographic location. Cloud computing may present a cloud to users by connecting resources at multiple geographic locations. By connecting resources at multiple geographic locations to organize a cloud, cloud computing may meet problems of communication interception, congestion, and interruption. Cloud computing should have a way to supply extra processing on demand for certain links between computers separated geographically. Since a MapReduce cloud is the key to the success of the large-scale computation, cloud computing can use the Smart MapReduce Cloud (SMRC) proposed in this paper to apply extra processing to intermediate data on demand while intermediate data is delivered among computers in the MapReduce cloud. In experiments, cloud computing is tested with several popular MapReduce applications to observe performances of data encryption and compression via XOR and GZIP functions in SMRC.

Smart Intermediate Data Transfer for MapReduce on Cloud Computing

MapReduce is a programming model proposed by Google to process large datasets in clusters. However, MapReduce often needs to transfer much intermediate data among nodes, which is harmful to performances of an application. MapReduce can be enhanced by using the proposed Smart Intermediate Data Transfer (SIDT) in the runtime system to smartly arrange intermediate data. Although SIDT does not reduce intermediate data to the minimal size in comparison with other intermediate data arrangement procedures such as Huffman coding, bzip2, and gzip, MapReduce is proved to get a better performance from SIDT than from others in the experiments of this paper.

Smart Task Distributor for MapReduce on Cloud Computing

A MapReduce system is widely used to implement the large-scale computation on cloud computing. A MapReduce system currently defines computation resources in a node as a roughly configurable slot number, and distributes tasks over nodes according to the slot number. However, a MapReduce system may make computation resources of clusters in the underutilization or overutilization condition, because a task of different applications unlikely uses the same computation resources and because a node may have different CPUs with different capabilities. A MapReduce system can use Smart Task Distributor (STD) proposed in this paper to solve the problem of the computation resource underutilization or overutilization in clusters. Technically, a MapReduce system can use STD to smartly distribute tasks over nodes in clusters, because STD on the one hand gradually assigns tasks to a node in order to fully utilize computation resources in the node and on the other hand dynamically estimates the remaining computation resources in the node for toggling the assignment of tasks on demand without overloading it. In experiments, a MapReduce system is proved to get better performances with STD than with other ways.

A power saving scheme for delivering non-real-time data in IEEE 802.16m

How to save power in wireless networks is a crucial problem because the official power saving scheme in IEEE 802.16m does not work efficiently. In this paper, we propose a power saving scheme for delivering non-real-time data. According to behaviors of a client in communication, our power saving scheme can dynamically adjust the sleep time of a client to avoid waking up the client and consuming its power. We use simulations to prove that a client can save more power with our power saving scheme than with the official power saving scheme in IEEE 802.16m.