We investigate the issues involved in developing a scalable World Wide Web (WWW) server on a cluster of workstations and parallel machines, using the Hypertext Transport Protocol (HTTP). The main objective is to strengthen the processing capabilities of such a server by utilizing the power of multicomputers to match huge demands in simultaneous access requests from the Internet. We have implemented a system called SWEB on a distributed memory machine, the Meiko CS-2, and networked SUN and DEC workstations. The scheduling component of the system actively monitors the usages of CPU, I/O channels and the interconnection network to effectively distribute HTTP requests across processing units to exploit task and I/O parallelism. We present the experimental results on the performance of this system. Our results indicate that the system delivers good performance on multi-computers and obtains significant improvements over other approaches. 1 Motivation The Scalable Web server (SWEB) project ...

The server array is a novel video server architecture based on partitioning each video over multiple server nodes, thereby achieving perfect load balancing for any demand distribution. We discuss the main design issues, compute the buffer requirements at the client, and compare the reliability of different video server architectures.

The performance of striped disk arrays is governed by two parameters: the stripe unit size and the degree of striping. In this paper, we describe techniques for determining the stripe unit size and degree of striping for disk arrays storing variable bit rate continuous media data. We present an analytical model that uses the server configuration and the workload characteristics to predict the load on the most heavily loaded disk in redundant and non-redundant arrays. We then use the model to determine the optimal stripe unit size for different workloads. We also use the model to study the effect of various system parameters on the optimal stripe unit size. To determine the degree of striping, we first demonstrate that striping a continuous media stream across all disks in the array causes the number of clients supported to increase sub-linearly with increase in the number of disks. To maximize the number of clients supported in large arrays, we propose a technique that partitions a disk array and stripes each media stream across a single partition. Since load imbalance can occur in such partitioned arrays, we present an analytical model to compute the imbalance across partitions in the array. We then use the model to determine a partition size that minimizes the load imbalance, and hence, maximizes the number of clients supported by the array.

In this paper, we investigate the issues involved in developing a scalable World Wide Web (WWW) server called SWEB on a cluster of workstations. The objective is to strengthen the processing capabilities of such a server in order to match huge demands in simultaneous access requests from the Internet, especially when these requests involve delivery of large digitized documents. The scheduling component of the system actively monitors the usages of CPU, disk I/O channels and the interconnection network to effectively distribute HTTP requests across processing units to exploit task and I/O parallelism. We analyze the maximum number of requests that can be handled by the system and present several experiments to examine the performance of this system. 1 Motivation The Scalable Web server (SWEB) project grew out of the needs of the Alexandria Digital Library (ADL) project at UCSB [A96]. Digital library systems, which provide the on-line retrieval and processing of digitized documents thro...

In this paper, we investigate the suitability of clustered architectures for designing scalable multimedia servers. Specifically, we evaluate the effects of: (i) architectural design of the cluster, (ii) the size of the unit of data interleaving, and (iii) read-ahead buffering and scheduling on the real-time performance guarantees provided by the server. To analyze the effects of these parameters, we develop an analytical model of clustered multimedia servers, and then validate it through extensive simulations. The results of our analysis have formed the basis of our prototype implementation based on a cluster of switch-connected RS/6000 machines. We briefly describe the prototype and discuss some implementation details. 1. Introduction A wide range of applications in entertainment, business, and education require the development of multimedia servers that can support efficient means of storing and delivering digital audio and video objects to thousands of clients. The fundamental pr...

The performance of striped disk arrays is governed by two parameters: the stripe unit size and the degree of striping. In this paper, we describe techniques for determining the stripe unit size and degree of striping for disk arrays storing variable bit rate continuous media data. We present an analytical model that uses the server configuration and the workload characteristics to predict the load on the most heavily loaded disk in redundant and non-redundant arrays. We then use the model to determine the optimal stripe unit size for different workloads. We also use the model to study the effect of various system parameters on the optimal stripe unit size. To determine the degree of striping, we first demonstrate that striping a continuous media stream across all disks in the array causes the number of clients supported to increase sub-linearly with increase in the number of disks. To maximize the number of clients supported in large arrays, we propose a technique that partitions a disk array and stripes each media stream across a single partition. Since load imbalance can occur in such partitioned arrays, we present an analytical model to compute the imbalance across partitions in the array. We then use the model to determine a partition size that minimizes the load imbalance, and hence, maximizes the number of clients supported by the array.

Mitra is a scalable storage manager that supports the display of continuous media data types, e.g., audio and video clips. It is a software based system that employs off-the-shelf hardware components. Its present hardware platform is a cluster of multi-disk workstations, connected using an ATM switch. Mitra supports the display of a mix of media types. To reduce the cost of storage, it supports a hierarchical organization of storage devices and stages the frequently accessed objects on the magnetic disks. For the number of displays to scale as a function of additional disks, Mitra employs staggered striping. It implements three strategies to maximize the number of simultaneous displays supported by each disk.

We report on the design, implementation and evaluation of a scalable real-time streaming architecture, termed Yima, that enable applications such as video-on-demand and distance learning on a large scale. While Yima incorporates lessons learned from first generation research prototypes, it also complies with industry standards in content format (MP4) and communication protocol (RTP/RTSP). Meanwhile, it distinguishes itself from both research prototypes and commercial products because it incorporates our recent research and it pushes the industry envelope by supporting HDTV and multi-channel panoramic clients. This article describes the principles of existing approaches (both research and commercial) and how Yima improves upon them through our research. We also included some detailed information on how we tweaked available hardware and software to achieve certain objectives (e.g., playback of HDTV streams on an HDTV monitor). We describe three major contributions of Yima. First, we show how we have reduced inter-nodal data exchange and hence improved the scalability of Yima using a bipartite design. We report our comparison results of two such designs, both implemented in a real-world experimental setup. Second, we explain two alternative approaches to handling variable-bit-rate (VBR) video. Finally, in order to recover from RTP’s potential packet-loss, we report on our observations in integrating a selective retransmission protocol into Yima’s RTP server. Since packet-loss can mainly be observed in a WAN environment, we experimented with streaming MPEG-4 data over the Internet. Yima is operational and supports a variety of display bandwidths from MPEG-4 at 800 Kb/s to five synchronized channels of MPEG-2 at a total of 20 Mb/s. 1

Abstract not found

. The growing number of digital audio and video repositories has resulted in a desperate need for effective techniques to deliver data to users in a timely manner. Due to geographical distribution of users, it is not cost effective to have a centralized media server. In this paper, we investigate issues involved in the design of a distributed video server (DVS) to support movie-on-demand (MOD) application. We propose a redundant hierarchical (RedHi) architecture for DVS where the nodes are continuous media servers and the edges are dedicated network lines. With RedHi, each node has two or more parents. We show that the redundant links in RedHi yield a more reliable and efficient system. Our simulation results demonstrate that RedHi can tolerate a single link failure with no degradation in performance while with pure hierarchy almost 2:5% of requests are rejected due to the failure. In normal mode of operation, RedHi outperforms pure hierarchy significantly (160% improvement on the aver...