Deploying a modern software service usually involves installing several software components, and configuring these components properly to realize the complex interdependencies between them. This process, which accounts for a significant portion of information technology (IT) cost, is complex and error-prone. In this paper, we propose delta configuration - an approach that reduces the cost of software deployment by eliminating a large number of choices on parameter values that administrators have to make during deployment. In delta configuration, the complex software stack of a distributed service is first installed and tested in a test environment. The resulting software images are then captured and used for deployment in production environments. To deploy a software service, we only need to copy these pre-configured software images into a production environment and modify them to account for the difference between the test environment and a production environment. We have implemented a prototype system that achieves software deployment using delta configuration of the configuration state captured inside virtual machines. We perform a case study to demonstrate that our scheme leads to substantial reduction in complexity for the customer, over the traditional software deployment method.

Given a set of machines and a set of Web applications with dynamically changing demands, an online application placement controller decides how many instances to run for each application and where to put them, while observing all kinds of resource constraints. This NP hard problem has real usage in commercial middleware products. Existing approximation algorithms for this problem can scale to at most a few hundred machines, and may produce placement solutions that are far from optimal when system resources are tight. In this paper, we propose a new algorithm that can produce within 30 seconds high-quality solutions for hard placement problems with thousands of machines and thousands of applications. This scalability is crucial for dynamic resource provisioning in large-scale enterprise data centers. Our algorithm allows multiple applications to share a single machine, and strives to maximize the total satisfied application demand, to minimize the number of application starts and stops, and to balance the load across machines. Compared with existing state-of-the-art algorithms, for systems with 100 machines or less, our algorithm is up to 134 times faster, reduces application starts and stops by up to 97%, and produces placement solutions that satisfy up to 25% more application demands. Our algorithm has been implemented and adopted in a leading commercial middleware product for managing the performance of Web applications.

Since many Internet applications employ a multitier architecture, in this article, we focus on the problem of analytically modeling the behavior of such applications. We present a model based on a network of queues where the queues represent different tiers of the application. Our model is sufficiently general to capture (i) the behavior of tiers with significantly different performance characteristics and (ii) application idiosyncrasies such as session-based workloads, tier replication, load imbalances across replicas, and caching at intermediate tiers. We validate our model using real multitier applications running on a Linux server cluster. Our experiments indicate that our model faithfully captures the performance of these applications for a number of workloads and configurations. Furthermore, our model successfully handles a comprehensive range of resource utilization---from 0 to near saturation for the CPU---for two separate tiers. For a variety of scenarios, including those with caching at one of the application tiers, the average response times predicted by our model were within the 95% confidence intervals of the observed average response times. Our experiments also demonstrate the utility of the model for dynamic capacity provisioning, performance prediction, bottleneck identification, and session policing. In one scenario, where the request arrival rate increased from less than 1500 to nearly 4200 requests/minute, a dynamic provisioning technique employing our model was able to maintain response time targets by increasing the capacity of two of the tiers by factors of 2 and 3.5, respectively.

We introduce and evaluate a middleware clustering technology capable of allocating resources to web applications through dynamic application instance placement. We define application instance placement as the problem of placing application instances on a given set of server machines to adjust the amount of resources available to applications in response to varying resource demands of application clusters. The objective is to maximize the amount of demand that may be satisfied using a configured placement. To limit the disturbance to the system caused by starting and stopping application instances, the placement algorithm attempts to minimize the number of placement changes. It also strives to keep resource utilization balanced across all server machines. Two types of resources are managed, one load-dependent and one load-independent. When putting the chosen placement in effect our controller schedules placement changes in a manner that limits the disruption to the system.

Web applications have assumed a central role in the enterprise data center, and are becoming the foundation of an increasing number of critical business functions. The availability of such applications is a problem of critical importance. To address this problem, we present a health monitoring and control system for application server environments. We build on past work in the area of measuring system health by evaluating a number of parameters that convey health information. We propose a class-based model that assigns health policies to application servers to monitor health conditions of interest and correct problems before they cause catastrophic damage. We describe an implementation of the system, and provide a utility-based experimental evaluation.

The deployment and configuration of distributed applications is a human intensive and highly complex process that poses significant challenges to data center operators. The process involves many cross-cutting concerns such as connectivity, performance, and security requirements, as well as resource availability, policies and best practices. These interdependencies represent a significant source of complexity, cost, and risk in data center management. In this paper we address this problem using a new approach that leverages concepts from the model-driven architecture research domain. We describe a prototype application deployment automation system based on model transformation techniques. We show how model transformation techniques can replace the manual process of writing and adapting scripts and workflows, reduce the deployment complexity, guarantee configuration integrity and consistency, and allow for a separation of concerns.

Since many Internet applications employ a multi-tier architecture, in this paper, we focus on the problem of analytically modeling the behavior of such applications. We present a model based on a network of queues, where the queues represent different tiers of the application. Our model is sufficiently general to capture (i) the behavior of tiers with significantly different performance characteristics and (ii) application idiosyncrasies such as session-based workloads, concurrency limits, and caching at intermediate tiers. We validate our model using real multi-tier applications running on a Linux server cluster. Our experiments indicate that our model faithfully captures the performance of these applications for a number of workloads and configurations. For a variety of scenarios, including those with caching at one of the application tiers, the average response times predicted by our model were within the 95% confidence intervals of the observed average response times. Our experiments also demonstrate the utility of the model for dynamic capacity provisioning, performance prediction, bottleneck identification, and session policing. In one scenario, where the request arrival rate increased from less than 1500 to nearly 4200 requests/min, a dynamic provisioning technique employing our model was able to maintain response time targets by increasing the capacity of two of the application tiers by factors of 2 and 3.5, respectively.

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. 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.

The objective of this study is investigate the performance difference of packet switching architectures working in a synchronous and asynchronous way respectively; in such architectures the packet contention is resolved in the wavelength domain and the used wavelength converters are shared; we investigate on the saving of the number of converters that the sharing technique allows to obtain in the synchronous and asynchronous architectures and compare the obtained results. These ones show that when a packet loss probability is fixed, in the synchronous case a greater number of converters is saved, in some cases the gain is 40% more than the asynchronous case; furthermore in the asynchronous case a more expensive switching matrix is needed.

We propose an architecture for hybrid services, i.e., services that span many network technologies, such as the public switched telephone network (PSTN), cellular networks, and networks based on IP. These services will play an important role in the future because they leverage on the existing infrastructures rather than requiring new and sophisticated mechanisms to be deployed. We explore a few issues related to hybrid services and propose a platform as well as a set of components to facilitate their creation and deployment. The existing infrastructure is only required to generate specific events when requests for hybrid services are detected. We present the design of a service layer, based on Java, that handles the treatment of these special requests. Our service layer is provided with a set of generic components realized according to the JavaBeans model. We illustrate the strength of our architecture by discussing two hybrid-service examples: a calendar service and a call forwarding service.

Outlines a methodology for developing network control systems which allows for an evaluation of the dynamic behavior and overall performance at an early stage of the development process. Our approach is to build a software prototype which is designed according to the architecture under consideration and runs the intended control algorithms. The functional and dynamic properties of this prototype are tested and evaluated on an emulation platform that we built for this purpose. By providing support for real-time visualization and interactive emulation, this platform can be used to study multimedia networks in various scenarios, such as different load patterns, network sizes and management operations. The current implementation runs on a KSR-1 and an SP2 parallel processor, which are connected to a graphics workstation via ATM links. We use the platform in several projects, one of which aims at developing an architecture for managing multimedia network services.

We propose a new policy for virtual path bandwidth allocation in broadband networks. Based on a threshold scheme, our policy handles the inherent tradeoff between bandwidth utilization and processing costs. In each virtual path controller the thresholds are chosen so as to keep bandwidth utilization high, while obtaining a low rate of processing requests. Two novel ideas are used in our threshold scheme: adaptivity, which results in a better prediction of future bandwidth requirements; and hysteresis, which prevents excessive processing of requests due to oscillations around thresholds. We tested the performance of our new bandwidth control scheme, and compared it with previously suggested schemes. The performance measures were the expected amount of unused bandwidth, the average signaling load and the blocking probability. Performance has been evaluated through numerical computations as well as by simulations. Our analysis is based on a time segmentation technique which allows us to reduce a Markov chain with NM states into M Markov chains with N states and a one-dimensional chain with M states. Our results show that our policy significantly improves upon previously suggested approaches.

No abstract available.

The authors review the results of a two-level control framework for integrated video and image traffic in an ATM multiplexer. The framework gives priority to video traffic, and uses flow control on the image traffic in order to prevent congestion while improving its throughput. It functions on two levels, or time scales. Buffer management and scheduling are done at the level of cell arrivals and departures, or microsecond time scales, while flow control is done at the burst level, or millisecond time scales. The control framework guarantees upper bounds on cell loss rates for both video and image traffic over a defined control interval, with a specified probability.>

A framework for joint scheduling and admission control in broadband switching systems based on a principle of separation between these two levels of control is developed. It is shown that an admission control strategy can be tailored to a particular mix of traffic by using high-level information from the scheduler. This principle is presented in the context of asynchronous time-sharing (ATS) in which explicit guarantees of cell-level and call-level quality of service (QOS) are given to several traffic classes. The separation principle allows the formulation of an optimal admission control policy that maximizes the expected system utility and maintains all QOS guarantees. Several heuristic admission control policies are considered and compared with the optimal policy. The admissible load region is introduced as a means of quantifying the capacity of a switch under the QOS constraints at the cell and call levels. Numerical calculations for a single MAGNET II switching node carrying two classes of real-time traffic are used to illustrate the effects of different scheduling and admission control policies on both the expected utility and the admissable load region.>

A proactive distributed cooperative control algorithm for resource scheduling of broadband networks based on asynchronous time sharing is presented. The problem setting for distributed, cooperative proactive scheduling is described. An overview of the distributed cooperative algorithm is presented. The general control principles underlying the effectiveness of the cooperative algorithm are discussed. The formal details of the distributed cooperative algorithm and its underlying single noise scheduling algorithm are considered. Simulation results comparing the distributed cooperative algorithm with other networkwide scheduling algorithms are reported.>

A joint scheduling and admission control algorithm is presented for Asynchronous Time-Sharing (ATS)-based switching nodes carrying real-time traffic. Systems based on ATS guarantee quality of service, at both the levels of cells and calls, for three well-defined traffic classes. A mechanism is outlined by which an admission control strategy can be tailored to a particular mix of traffic classes. A mechanism is outlined by which an admission control strategy can be tailored to a particular mix of traffic by making use of high-level information from the scheduler. This mechanism is based on a principle of separation between scheduling and admission control. A linear programming formulation is used to find the admission control policy which will maximize the expected system utility while maintaining the guaranteed quality of service. The admissible load region is introduced as a means of quantifying the capacity of a switch under the QOS constraints at the cell and call levels. Numerical calculations for a single MAGNET II switching node carrying two classes of real-time traffic are used to illustrate the effects of different scheduling and admisssion control policies on both the expected utility and the admissible load region.

A real-time scheduling algorithm for Asynchronous Time Sharing based switching nodes is presented.Three classes of traffic sources are considered to load the swit thing nodes: real-time video sources with guaranteed maximum end-to-end delay S1 for all cells; realtime voice sources with guaranteed end-to-end delay SII, c% cell loss rate and average gap length q; and data traffic with guaranteed minimum average throughput and maximum average time delay.The proposed algorithm is compared with other known algorithms in terms of complexity in implementation, knowledge structures and performance.Our investigations are supported throughout by simulation results.

Whether or not the introduction of traffic classes improves upon the performance of ATM networks is discussed within the framework provided by a class of networks that guarantees quality of service. To provide a meaningful comparison the authors define the concept of a schedulable region, a region in the space of loads for which the quality of service is guaranteed. The authors show the dependence of the schedulable region on the scheduling algorithm employed, quality of service parameters, and traffic statistics. An efficient real-time scheduling algorithm is introduced that substantially increases the schedulable region without incurring prohibitive complexity costs. The schedulable region associated with this algorithm is compared with the ones generated by the static priority scheduling algorithm and a variant of the minimum laxity threshold algorithm. The size and shape of the schedulable region is explored by means of simulations.>

An access protocol suitable for asynchronous transfer mode (ATM) operation in an integrated service broadband metropolitan area network (MAN) is presented. This protocol is designed for a twin-bus optical network and is characterized by a fully distributed control mechanism. A detailed description of the MAC procedure and the results of its performance evaluation are given. In particular, the high fairness degree and the very high throughput values that are the distinctive features of this protocol are highlighted and compared with those relevant to the current version of the IEEE standard 802.6.>

Real-time traffic measurements on MAGNET II, an integrated network testbed based on asynchronous time sharing, are reported. The quality of service is evaluated by monitoring the buffer-occupancy distribution, the packet time-delay distribution, the packet loss, and the gap distribution of the consecutively lost packets. The experiments show that both time-delay and buffer-occupancy distributions of multiplexed video sources display a marked bimodal behavior, which does not seem to depend on the buffer size. The reliance of the network designer on traffic sources that do not exhibit substantial correlations can lead to implementations with serious congestion problems. For asynchronous-time-sharing-based networks with different traffic classes, the impact of a traffic class on the performance of the other classes tends to be diminished when compared to single-class-based asynchronous transfer mode (ATM) networks.>