No abstract

We present an architecture and prototype implementation of a performance management system for cluster-based web services. The system supports multiple classes of web services traffic and allocates server resources dynamically so to maximize the expected value of a given cluster utility function in the face of fluctuating loads. The cluster utility is a function of the performance delivered to the various classes, and this leads to differentiated service. In this paper we will use the average response time as the performance metric. The management system is transparent: it requires no changes in the client code, the server code, or the network interface between them. The system performs three performance management tasks: resource allocation, load balancing, and server overload protection. We use two nested levels of management mechanism. The inner level centers on queuing and scheduling of request messages. The outer level is a feedback control loop that periodically adjusts the scheduling weights and server allocations of the inner level. The feedback controller is based on an approximate first-principles model of the system, with parameters derived from continuous monitoring. We focus on SOAP-based web services. We report experimental results that show the dynamic behavior of the system.

We present an architecture and prototype implementation of a performance management system for cluster-based web services. The system supports multiple classes of web services traffic and allocates server resonrces dynamically so to maximize the expected value of a given cluster utility function in the face of fluctuating loads. The cluster utility is a function of the performance delivered to the various classes, and this leads to differentiated service. In this paper we will use the average response time as the performance metric. The management system is Iransparenf: it requires no changes in the client code, the server code, or the network interface between them. The system performs three performance management tasks: resource allocation, load balancing, and server overload protection. We use two nested levels of management mechanism. The inner level centers on queuing and scheduling of request messages. The outer level is a feedback control loop that periodically adjusts the scheduling weights and server allocations of the inner level. The feedback controller is based on an approximate first-principles model of the system, with parameters derived from continuous monitoring. We focus on SOAP-based web services. We report experimental results that show the dynamic behavior of the system.

We present an architecture and prototype implementation of a performance management system for cluster-based web services. The system supports multiple classes of web services traffic and allocates server resources dynamically so to maximize the expected value of a given cluster utility function in the face of fluctuating loads. The cluster utility is a function of the performance delivered to the various classes, and this leads to differentiated service. In this paper we will use the average response time as the performance metric. The management system is transparent: it requires no changes in the client code, the server code, or the network interface between them. The system performs three performance management tasks: resource allocation, load balancing, and server overload protection. We use two nested levels of management mechanism. The inner level centers on queuing and scheduling of request messages. The outer level is a feedback control loop that periodically adjusts the scheduling weights and server allocations of the inner level. The feedback controller is based on an approximate first-principles model of the system, with parameters derived from continuous monitoring. We focus on SOAP-based web services. We report experimental results that show the dynamic behavior of the system.

Distance learning-education without a central classroom-has helped busy people obtain college credits or complete training they might otherwise not have done. Methods of distance learning range from simple correspondence courses and broadcast TV with reverse audio to specialized video conferencing tools, such as Proshare or Flashback, and Web-based courses. The current version of Old Dominion University's Teletechnet system, for example, uses broadcast satellite technology with terrestrial audio feedback from students and e-mail to connect the main campus in Norfolk, Virginia, to up to 23 community colleges throughout the state, as well as selected industrial and government sites. More than 3,000 students are enrolled in Teletechnet. However, limitations in the technologies supporting Teletechnet and similar systems become critically apparent as the demand for them continues to rise. To address these limitations, our research group built the Interactive Remote Instruction system, which allows students at geographically dispersed satellite campuses and community colleges to take a class "together." Access from home PCs through a Windows NT port is planned but not yet available. IRI improves on Teletechnet technology in four areas: ♦ Video resolution. The limited resolution of Teletechnet's TV images restricts the quality of information that can be presented. IRI offers images with a resolution of 1152 x 900 pixels. ♦ Asymmetrical video presence. Instructors cannot view the students at remote Teletechnet sites nor can students see students at other sites when they speak. With IRI, each course participant's (student's and instructor's) workstation is a window to a virtual classroom. All class participants see the speakers on their screens, so everyone has the same experience. ♦ Interaction. Teletechnet students have limited opportunities to interact both in and out of class. In class, IRI helps each student prepare for assignments and take notes and aids collaboration in group projects. ♦ Instructor support. TV-based classes typically have technicians and camera people managing the connection but little user support is provided. IRI helps instructors prepare a lesson plan off-line, which they can then use to guide class presentations. The class management helps orchestrate and manage an interactive classroom. , The instructor can selectively call on students or check the status of their workstations. ♦ Computer simulations. TV-based systems provide no way to do a hands-on computer simulation of, say, a chemical experiment. With IRI, students can ask questions or demonstrate proficiency by directly manipulating such a simulation. In this article we describe IRI and the lessons learned deploying it. We first deployed IRI to teach a fall 1995 graduate course in software metrics. We then evaluated it in terms of logistics, reliability, performance, and usability, performing off-line experiments to try out new features and to develop protocols that would improve IRI use. We subsequently reengineered IRI into an open arch...