Abstract: Multimedia is gaining ever more importance on the Internet. This increases the need for intelligent and efficient video caches. Typical Web proxies were not designed to efficiently support the caching of videos. A promising approach to improve caching efficiency is to adapt videos. With the availability of MPEG-4 it is possible to develop a standard compliant proxy that allows fast and efficient adaptation. We propose a modular design for an adaptive MPEG-4 video proxy that supports efficient full and partial video caching in combination with filtering options that are driven by the terminal capabilities of the client. We use the native scalability operations provided by MPEG-4 and use the emerging MPEG-7 standard to describe the scalability options for a video. The proxy parses the MPEG-7 description and decides, based on this description and the terminal capabilities of the client, which adaptation step to choose. Simple MPEG-4 audio-visual streams are supported by filter operations in the compressed domain that realize several temporal scaling algorithms and color reduction. In this paper, we will restrict ourselves to full video caching. The combination of adaptation with MPEG-4, MPEG-7 and client terminal capabilities is to the best of our knowledge unique and will increase the quality of service for end users.

Abstract: Multimedia streaming becomes ever more popular. The multimedia standard MPEG-4 has been designed to support scenes of different levels of complexity and applications with low bandwidth requirements up to very high bandwidth requirements. One protocol suitable to transfer this kind of data over IP networks is the real-time transport protocol (RTP). This report describes standardized and proposed payload formats that support the transport of MPEG-4 elementary streams over RTP connections. These RTP packetization formats are compared w.r.t. their suitability for the adaptation (scaling) of the media data within the network, i.e., by advanced routers or proxy caches. This adaptation process is governed by metadata that need to be transferred and inspected in conjunction with the media streams.

Abstract: The virtual interface (VI) architecture standard was developed to satisfy the need for a high throughput, low latency communication system required for cluster computing. VI architecture aims to close the performance gap between the bandwidths and latencies provided by the communication hardware and visible to the application, respectively, by minimizing the software overhead on the critical path of the communication. This paper presents the results of a performance study of one VI architecture hardware implementation, the Giganet cLAN (cluster LAN). The focus of the study is to assess and compare the performance of different VI architecture data transfer modes and specific features that are available to higher-level communication software like MPI in order to aid the implementor to decide which VI architecture options to employ for various communication scenarios. Examples of such options include the use of send/receive vs. RDMA data transfers, polling vs. blocking to check completion of communication operations, multiple VIs, completion queues and scatter capabilities of VI architecture.

Abstract: In this paper we present the design, architecture and implementation of the A2QM3 System. It provides programmers re-usable QoS-aware Control Objects to enable building a complete middleware for adaptive applications over active networks. We introduce the programming model, the system architecture, and show the parts that make this system a full-featured middleware supporting QoS-aware reliable stream-oriented communication, communication using the request/reply-based CORBA model and real-time streaming for continuous multimedia contents.

Abstract: Adaptation is becoming an increasingly important tool for resource and media management in distributed multimedia systems. Best-effort scheduling and worst-case reservation of resources are two extreme cases, none of them well suited to cope with large-scale, dynamic multimedia systems. The middle course can be met by a system which dynamically adapts its data, resource requirements, and processing components to achieve user satisfaction. Nevertheless, there is no agreement about the questions, where, when, what and who should adapt. A number of papers have been published in recent years, where adaptation is a central issue, however, in most different interpretations and generally in a somehow limited scope; e.g.,[1, 2, 8, 9, 10, 12]. A distributed multimedia system comprises several types of components, such as media servers, meta-databases, proxies, routers, clients. Also, a large number of adaptation possibilities exist, from simple frame dropping up to virtual server systems which dynamically allocate new resources on demand. The main question is, which kind of component can be best used for what kind of adaptation. In the ADMITS project (Adaptation in Distributed Multimedia IT Systems), we are seeking for answers to exactly this basic question, and to a number of related questions.

Abstract: This paper presents an inexpensive cluster-based QoS networking testbed that can be employed to \emulate" different networks for multimedia communication ex-periments. Such a network can be built using standard PC and Ethernet hardware and open-source software components, e.g., IP routing and traÆc control avail- able in recent Linux kernels as well as a Differentiated Services package built atop these building blocks. The testbed can exibly be configured to model various link bandwidths as well as IP routers capable of classifying, queuing (with various disciplines), forward-ing and/or dropping packets and shaping traÆc. The QoS components and facilities of the testbed are in-troduced and initial performance analysis experiments and results are reported. A simple video streaming application under QoS control is presented to show the usefulness of the testbed.

Abstract: Current superscalar microprocessors are able to operate at a peak performance of up to 1 GFlop/sec. However, current main memory technology does not provide the data needed fast enough to keep the CPU busy. To minimize idle times of the CPU, caches are used to speed up accesses to frequently used data. To exploit caches, the software must be aware of them and reuse data in the cache before it is being replaced. Unfortunately, all conventional multigrid codes are not cache-aware and hence exploit less than 10 percent of the peak performance of cache based machines. Our studies with linear PDEs with constant coefficients show that it is possible to speed up the execution of our multigrid method by a large factor and hence solve a Poisson’s equation with one million unknowns in less than 3 seconds. The optimized reuse of data in the cache allows us to exploit 30 percent of the peak performance of the CPU, in contrast to mgd9v for instance, which achieves less than 5 percent on the same machine. To achieve this, we used several techniques like loop unrolling and loop fusion to better exploit the memory hierarchy and the superscalar CPU. We study the effects of these techniques on the runtime performance in detail. We also study several tools which guide the optimizations and help to restructure the code.

Abstract: This paper describes an unconventional way to apply a performance analysis tool for parallel programs (Vampir) to understand and tune the performance of the real-time voice and data communication software running on top of Frequentis’ V4 switch. The execution schedule of the strictly time-triggered V4 switching software is computed off-line; analyzing the schedule to identify e.g. performance bottlenecks used to be a complex and time-consuming process. We present our approach to transform the V4 software schedule’s information into Vampir trace files and use this tool’s facilities to provide a visualization of the schedule. A case study illustrates the benefits of this approach.

Abstract: The Virtual Interface (VI) Architecture standard was developed to satisfy the need for a high-throughput, low-latency communication system required for cluster computing. This paper presents the results of a performance study of one VI Architecture hardware implementation, the Giganet cLAN (Cluster LAN). The focus of the study is to assess and compare the performance of different VI Architecture data transfer modes and specific features that are available to higher-level communication software like MPI, in order to aid the implementor to decide which VI Architecture options to employ for various communication scenarios. Examples of such options include the use of send/receive vs. RDMA data transfers, polling vs. blocking to check completion of communication operations, multiple VIs, completion queues, and scatter capabilities of VI Architecture.

Abstract: Workstation and PC clusters interconnected by SCI (Scalable Coherent Interface) are very promising technologies for high performance cluster computing. Using commercial SBus to SCI interface cards and early system software and drivers, a two-workstation cluster has been constructed for initial testing and evaluation. The PVM system has been adapted to operate on this cluster using raw device access to the SCI interconnect, and preliminary communications performance tests have been carried out. Our preliminary results indicate that communications throughput in the range of 3.5 MBytes/s, and latencies Research supported by the Applied Mathematical Sciences program, Office of Basic Energy Sciences, U. S. Department of Energy, under Grant No. DE-FG05-91ER25105, the National Science Foundation, under Award Nos. ASC-9527186 and ASC-9214149, and the German Science Foundation SFB342. of 620 ¯s can be achieved on SCI clusters. These figures are significantly better (by a factor of 3 to 4) ...

Abstract: This paper describes the design of a common message passing layer for implementing both MPI and PVM over the SCI interconnect in a workstation or PC cluster. The design is focused at obtaining low latency. The message layer encapsulates all necessary knowledge of the underlying interconnect and operating system. Yet, we claim that it can be used to implement such different message passing libraries as MPI and PVM without sacrificing efficiency. Initial results obtained from using the message layer in SCI clusters are presented.

Abstract: Dieses Papier gibt eine allgemeine Einführung in Threads und vergleicht einige derzeit für Arbeitsplatzrechner erhältliche Thread-Systeme. Aufbauend auf einer Motivation und grundlegenden Erläuterung des Thread-Konzepts werden wichtige Aspekte und Probleme von Thread-Bibliotheken vorgestellt. Nach einigen Hinweisen zur Programmierung mit Threads werden mehrere Implementierungen einander gegenübergestellt.

Abstract: Due to their excellent cost/performance ratio, clusters of PCs can be attractive high-performance computing (HPC) platforms. Yet, their limited communication performance over standard LANs is still prohibitive for parallel applications. The project "Shared Memory in a LAN-like Environment" (SMiLE) at LRR-TUM adopts Scalable Coherent Interface (SCI) interconnect technology to build, and provide software for, a PC cluster which, with hardware-based distributed shared memory (DSM) and high-performance communication characteristics, is regarded as well suited for HPC. The paper describes the key features of the enabling technology, SCI. It then discusses the developments and important results of the SMiLE project so far: the development and initial performance of a PCI/SCI interface card, and the design and initial performance results of low-latency communication layers, Active Messages and a sockets emulation library.

Abstract: Low latencies for small messages are an important factor of efficient fine-grained parallel computation. The Active Messages concept provides this minimal overhead by eliminating certain parts of the critical path of sending and receiving messages, that is the context switch into the operating system kernel when using user-mode I/O, and multiple buffering in the network layer. Hardware-supported distributed shared memory (DSM) architectures exhibit various properties that make them particularly useful for an implementation of the aforementioned messaging mechanisms. This paper thus describes the concept, implementation, and the performance of a DSM-based Active Messages layer.

Abstract: To enable the efficient utilization of clusters of workstations it is crucial to develop a stable and rich software infrastructure. The ESPRIT Project SISCI will provide two widely used message-passing interfaces, MPI and PVM, as well as a POSIX compliant, distributed thread package (Pthreads) on multiple SCI-based clusters. This paper features motivation and background on this projects as well as details of the two core components: the common messaging layer and the Pthreads package.

Abstract: Active Messages haben sich als effizientes Kommunikationsverfahren insbesondere auf Kommunikationstechnologien durchgesetzt, die einen direkten Zugriff des Benutzers ohne Intervention des Betriebssystems zulassen. Als Nachteil der leichtgewichtigen Kommunikation erwiesen sich jedoch die nicht ausreichenden Schutzmechanismen, vor allem bei der Verwendung mehrerer Prozesse, die sich gleichzeitig einer Active-Message-Bibliothek bedienen. Die Spezifikation 2.0 der Berkeley-Active-Messages unternimmt nun den Versuch, f¨ur das bekannte und schnelle Kommunikationsverfahren Schutzabstraktionen vorzusehen. Im Rahmen dieser Arbeit wird die Implementation eines solchen Active-Message-Layers der Version 2.0 auf einem Cluster von SCI-gekoppelten Arbeitsplatzrechnern beschrieben. Wir k¨onnen zeigen, daß die zus¨atzlichen Schutzmechanismen nur wenig Einfluß auf die Leistung haben und somit der Vorteil der Active Messages, die leichtgewichtige, feingranulare Kommunikation, erhalten bleibt.