Ensembles of distributed, heterogeneous resources, also known as computational grids, have emerged as critical platforms for high-performance and resource-intensive applications. Such platforms provide the potential for applications to aggregate enormous bandwidth, computational power, memory, secondary storage, and other resources during a single execution. However, achieving this performance potential in dynamic, heterogeneous environments is challenging. Recent experience with distributed applications indicates that adaptivity is fundamental to achieving application performance in dynamic grid environments. The AppLeS (Application Level Scheduling) project provides a methodology, application software, and software environments for adaptively scheduling and deploying applications in heterogeneous, multiuser grid environments. We discuss the AppLeS project and outline our findings.

Adaptive computing on the Grid using AppLeS

Task Matching and Scheduling in Heterogeneous Computing Environments Using a Genetic-Algorithm-Based Approach

Methods and apparatuses provide real-time or near real-time streaming of content, specified in one or more playlists, using transfer protocols such as an HTTP compliant protocol. In one embodiment, a method can execute a user application on a client device to present media files and to control presentation of the media files. The method can further run a media serving process on the client device to retrieve a playlist specifying the media files and a media source at which the media files are available, to retrieve the media files from the media source, and to decode the media files retrieved. The media serving process can call the user application to process a custom URL in order to obtain an object referred to by the custom URL.

Real-time or near real-time streaming

Methods and apparatuses for real-time or near real-time streaming of content using transfer protocols such as an HTTP compliant protocol. In one embodiment, a method includes dividing a stream of data, representing the contiguous time based content of a program (e.g. a live video broadcast), into a plurality of distinct media files, and generating a playlist file having a plurality of tags and Universal Resource Indicators (URIs) indicating an order of presentation of the plurality of distinct media files. The plurality of media files and the playlist file can be made available for transmission to a client device which can retrieve the media files using the playlist file.

Playlists for real-time or near real-time streaming

While it is often suggested that moderate-scale ad hoc sensor networks are a promising approach to solving real-world problems, most evaluations of sensor network protocols have focused on simulation, rather than realworld, experiments. In addition, most experimental results have been obtained in limited scale. This paper describes a practical application of moderate-scale ad hoc sensor networks. We explore several techniques for reducing packet loss, including quality-based routing and passive acknowledgment, and present an empirical evaluation of the effect of these techniques on packet loss and data freshness.

https://www.comp.nus.edu.sg/~bleong/geographic/related/yarvis02realworld.pdf

Real-world experiences with an interactive ad hoc sensor network

The authors present the design of an adaptive service for streaming MPEG video over a best-effort IP network environment. The adaptive video streaming service exploits the inherent adaptiveness of video applications to perform controlled and graceful adjustments to the perceptual quality of the displayed MPEG video stream in response to fluctuations in the QoS delivered by the three primary components in the end-to-end path of a video stream, i.e., the video server, the network, and the receiver. The design supports multiple video adaptation techniques that can be applied individually or in combination to adapt the transmitted video stream in response to fluctuations in the QoS provided by the underlying system. A novel aspect of the design is a video adaptation algorithm that selects the adaptation delivering the best perceptual quality for the video playback for a given QoS delivered by the underlying infrastructure. They describe a prototype implementation of the adaptive video streaming service and present the results of a performance evaluation of this prototype system.

/pdf/adaptive-streaming-of-mpeg-video-over-ip-networks-47sbs2ot78.pdf

Adaptive streaming of MPEG video over IP networks

Conventional routing protocols used in wireline and wireless networks are not designed to operate in an autonomous (or ad hoc) wireless local area network (AWLAN) environment with asymmetric (i.e. unidirectional) links. This paper presents the design of a routing protocol called source-initiated adaptive routing algorithm (SARA), for AWLAN environments that may contain asymmetric links. To allow efficient operation in a symmetric network while ensuring correct operation in asymmetric networks, SARA supports two different modes of protocol operation: one optimized for symmetric networks and the other providing additional capabilities needed for operation in asymmetric networks at a higher cost. The ability of SARA to adapt its mode of operation and optimize its execution for each network environment is one of its major advantages and distinguishes it from existing routing protocols for, wireless networks. Furthermore, within each of its operating modes, SARA supports best-effort as well as policy-based routing.

Source-initiated adaptive routing algorithm (SARA) for autonomous wireless local area networks

Heterogeneous computing covers a great variety of situations. This study focuses on a particular application domain (iterative automatic target recognition tasks) and an associated specific class of dedicated heterogeneous hardware platforms. The contribution of this paper is that, for the computational environment considered, it presents a methodology for real-time on-line input-data dependent remappings of the application subtasks to the processors in the heterogeneous hardware platform using previously stored off-line statically determined mappings. That is, the operating system will be able to decide during the execution of the application whether or not to perform a remapping based on information generated by the application from its input data. If the decision is to remap, the operating system will be able to select a previously derived and stored mapping that is appropriate for the given state of the application (e.g., the number of objects it is currently tracking).

On-line use of off-line derived mappings for iterative automatic target recognition tasks and a particular class of hardware platforms

A major impediment to the development of networked video applications is the lack of adequate support for video multicast within the Internet Protocol (IP). This paper presents an approach for extending the capability of IP to support multicast of video streams in a heterogeneous networking environment where there are static variations in the capacities of the links and the video quality needs of the receivers as well as dynamic variations in the available bandwidth of the network links. The approach, called IP extensions for video multicast (IPEVM), uses an innovative traffic control mechanism within the network routers that dynamically adapts the quality of the video to changes in the available network bandwidth while ensuring fair treatment of all traffic contending for the network's best effort service. It also permits an application-selectable mix of resource reservation and adaptive bandwidth management to be used in building networked video applications.

Traffic control mechanism to support video multicast over IP networks

One of the major impediments to the widespread use of large-scale, distributed memory multiprocessors is the difficulty of efficiently partitioning and mapping application algorithms onto these machines so as to extract a large portion of the machines' peak performance. In this paper, we present the preliminary accomplishments of an ongoing effort aimed at automating the complex tasks of software partitioning and mapping during the: system definition phase of application development for distributed memory multiprocessors. We describe a technique called the Augmented Task Dependency Graph (ATDG) for representing the high-level design of the application software. The ATDG allows one to express functional parallelism as well as data parallelism in a manner that facilitates automated partitioning and mapping. We propose a new strategy for searching through the possible space of design choices for partitioning and mapping. The proposed approach, called hierarchical hybrid search, organizes the search space as a hierarchy of sub-spaces. It permits the use of different search techniques for searching through different search sub-spaces. Examples of search techniques that could be employed in the proposed approach include hill-climbing, simulated annealing, and genetic algorithms.

R.S. Ramanujan

Papers

Adaptive streaming of MPEG video over IP networks

Source-initiated adaptive routing algorithm (SARA) for autonomous wireless local area networks

On-line use of off-line derived mappings for iterative automatic target recognition tasks and a particular class of hardware platforms

Traffic control mechanism to support video multicast over IP networks

A framework for automated software partitioning and mapping for distributed multiprocessors