Web hosting providers are increasingly looking into dynamic hosting to reduce costs and improve the performance of their platforms. Instead of provisioning fixed resources to each customer, dynamic hosting maintains a variable number of application instances to satisfy current demand. While existing research in this area has mostly focused on the algorithms that decide on the number and location of application instances, we address the problem of efficient enactment of these decisions once they are made. We propose a new approach to application placement and experimentally show that it dramatically reduces the cost of application placement, which in turn improves the end-to-end agility of the hosting platform in reacting to demand changes.

Web server scalability can be greatly enhanced via hybrid data dissemination methods that use both unicast and multicast. Hybrid data dissemination is particularly promising due to the development of effective end-to-end multicast methods and tools. Hybrid data dissemination critically relies on document selection which determines the data transfer method that is most appropriate for each data item. In this paper, we study document selection with a special focus on actual end-point implementations and Internet network conditions. We individuate special challenges such as scalable and robust popularity estimation, appropriate classification of hot and cold documents, and unpopular large documents. We propose solutions to these problems, integrate them in MBDD (middleware support multicast-based data dissemination) and evaluate them on PlanetLab with collected traces. Results show that the multicast server can effectively adapt to dynamic environments and is substantially more scalable than traditional Web servers. Our work is a significant contribution to building practical hybrid data dissemination services.

Actuator networks will enable an unprecedented degree of distributed control of physical environments, and further progress will critically depend on the availability of a simulation platform that can capture both the physical and the communication dynamics.

Abstract — Sensing, actuation, and decision units can control a remote physical environment and enable physical actions regardless of distance. However, the effectiveness of networked control depends on its ability to tolerate network non-determinism, which in turn can be enhanced by the use of play-back buffers. Although play-back has been intensively studied in multi-media applications, play-back schemes differ significantly in networked control, which is characterized by different performance metrics and a different sequence of communication events. The primary contribution of this paper is an end-to-end algorithm that integrates play-back buffering, sensor sampling, and control. The algorithm is extensively validated on simulations and realtime wide-area emulations. The integrated algorithm canceled the effect of disturbances as much as a proportional controller under ideal network conditions. I.

In this paper, we propose a bandwidth allocation scheme for networked control systems that have their control loops closed over a geographically distributed network. We first formulate the bandwidth allocation as a convex optimization problem. We then present an allocation scheme that solves this optimization problem in a fully distributed manner. In addition to being fully distributed, the proposed scheme is asynchronous, scalable, dynamic and flexible. We further discuss mechanisms to enhance the performance of the allocation scheme. We present analytical and simulation results

A major problem in mobile web applications as well as the wireless Internet is the scalable delivery of data. The most popular solution for this problem is a hybrid system that uses broadcast push to scalably deliver the most popular data, and reserves broadcast pull for delivery of less popular data. Such a hybrid scheme introduces a variety of data management problems at the broadcast server. In this paper, we examine three of these problems: the push popularity problem, the document classification problem, and the bandwidth division problem. The push popularity problem is to estimate the popularity of the documents in the web site. The document classification problem is to determine which documents should be pushed and which documents must be pulled. The bandwidth division problem is to determine how much of the server bandwidth to devote to pushed documents and how much of the server bandwidth should be reserved for pulled documents. We propose simple and elegant solutions for these problems. We report on experiments with our system that validate our algorithms.

A novel broadcast technique for wormhole-routed parallel computers based on recursion is presented in this paper. It works by partitioning the interconnection graph into a number of higher-level subgraphs. Then, we identify the transmission subgraph (TSG) in each subgraph. Both the higher-level subgraphs and the TSGs are recursively defined, i.e., we split each level i subgraph into several level i+1 subgraphs and identify-level i+1 TSGs accordingly. We first split and scatter the source message into the TSG of the original graph. Next, in each recursive round message transmissions are from lower-level TSGs to higher-level TSGs and all transmissions at the same level happen concurrently. The algorithm proceeds recursively from lower-level subgraphs to higher level subgraphs until each highest-level subgraph (a single node) gets the complete message. We have applied this general paradigm to a number of topologies including two or higher dimension mesh/torus and hypercube. Our results show considerable improvements over all other algorithms for a wide range of message sizes under both one-port and all-port models.

Summary form only given. Multicast is an effective and scalable solution to disseminate data in the Internet. Middleware support for multicast-based data dissemination (MSMDD) aims to integrate state-of-the-art data management methods and multicast communication techniques and provide a scalable multicast-based data management layer to applications. By using multicast in MSMDD, the server can disseminate hot documents effectively. To evaluate multicast performance, previous work often utilizes network-level metrics. We employ application level criteria to analyze the multicast push performance in MSMDD. When packets are not lost on links, our result shows that the application-perceived (or client-perceived) performance is independent of the delay from the server to clients; the multicast rate is a major factor for the performance; an end-to-end multicast system achieves almost the same application-perceived performance as IP multicast under the same multicast rates. However, IP multicast can tolerate much more packet losses than the end-to-end multicast system.

This paper elaborates on a paradigm for quality-of-service (QoS) that is local, i.e., it does not depend on multinode cooperation. In order to maintain short queuing delays, we individuate flows that occupy a large fraction of a buffer and segregate those flows into a separate queue. The algorithm is provably fair and can avoid all packet reorderings. We show through extensive simulations that state requirements are minimal, and that other flows will benefit from short queuing delays while aggressive flows can still maintain high throughput.

A major problem in web database applications and on the Internet in general is the scalable delivery of data. One proposed solution for this problem is a hybrid system that uses multicast push to scalably deliver the most popular data, and reserves traditional unicast pull for delivery of less popular data. However, such a hybrid scheme introduces a variety of data management problems at the server. In this paper we examine three of these problems: the push popularity problem, the document classification problem, and the bandwidth division problem. The push popularity problem is to estimate the popularity of the documents in the web site. The document classification problem is to determine which documents should be pushed and which documents must be pulled. The band-width division problem is to determine how much of the server bandwidth to devote to pushed documents and how much of the server bandwidth should be reserved for pulled documents. We propose simple and elegant solutions for these problems. We report on experiments with our system that validate our algorithms.

A major problem on the Internet is the scalable dissemination of information. This problem is particularly acute exactly at the time when the scalability of data delivery is most important. One proposed solution to this scalability problem is to use multicast communication. However, allowing multicast communication introduces many nontrivial data management problems, such as caching, consistency, and scheduling. We have built a middleware that unifies and extends state-of-the-art data management methods and algorithms into one software distribution. Its flexible and extensible architecture is built from individual components that can be selected or replaced depending on the underlying multicast transport mechanism or on the application needs. Particular care has gone into the design of the algorithms to optimize the user-perceived level of service. We demonstrate our middleware within the context of the RODS application.

Describes an experimental exploration in Internet-based control of robots. The motivation of this work is that Internet communications can be exploited to achieve greater productivity from machines with local intelligence. Local intelligent systems contact human experts to solicit advice when a problem facing the machine is beyond its cognitive capabilities. This topic is explored in the context of a robot performing a sample domestic task (sorting laundry). An experimental system was constructed that has limited autonomous competence, but which proved to be significantly more productive through the use of occasional Internet-based human supervision.

Advances in wireless and optical communication, as well as in Internet multicast protocols, make broadcast and multicast methods an effective solution to disseminate data. In particular, repetitive server-initiated broadcast is an effective technique in wireless systems and is a scalable solution to relieve Internet hot spots. A critical issue for the performance of multicast data dissemination is the multicast schedule. Previous work focused on a model where each data item is requested by clients with a certain probability that is independent of past accesses. We consider the more complex scenario where a client accesses pages in blocks (e.g., a HTML file and all its embedded images), thereby introducing dependencies in the pattern of accesses to data. We present a sequence of heuristics that exploit page access dependencies. We measured the resulting client-perceived delay on multiple Web server traces, and observed an average speed-up over previous methods ranging from 8% to 91%. We conclude that scheduling for multi-item requests is a critical factor for the performance of repetitive broadcast.

Broadcast disks are an emerging paradigm for massive data dissemination. In a broadcast disk, data is divided into n equal-sized pages, and pages are broadcast in a round-robin fashion by a server. Broadcast disks are effective because many clients can simultaneously retrieve any transmitted data. Paging is used by the clients to improve performance, much as in virtual memory systems. However, paging on broadcast disks differs from virtual memory paging in at least two fundamental aspects: A page fault in the broadcast disk model has a variable cost that depends on the requested page as well as the current state of the broadcast. Prefetching is both natural and a provably essential mechanism for achieving significantly better competitive ratios in broadcast disk paging. In this paper, we design a deterministic algorithm that uses prefetching to achieve an O(n log k) competitive ratio for the broadcast disk paging problem, where k denotes the size of the client's cache. We also show a matching lower bound of $\Omega(n\log k)$ that applies even when the adversary is not allowed to use prefetching. In contrast, we show that when prefetching is not allowed, no deterministic online algorithm can achieve a competitive ratio better than $\Omega(nk)$. Moreover, we show a lower bound of $\Omega(n \log k)$ on the competitive ratio achievable by any nonprefetching randomized algorithm against an oblivious adversary. These lower bounds are trivially matched from above by known results about deterministic and randomized marking algorithms for paging. An interpretation of our results is that in the broadcast disk paging, prefetching is a perfect substitute for randomization.

Broadcast disks are an emerging paradigm for mastive information dissemination.In a broadcast disk, data is divided into n equal-sized pages and pages are broadcast in a roundrobin manner by a server.Broadcast disks are effective because many clients can simultaneously retrieve any transmitted information.Paging is used by the clients to improve performance, much as in virtual memory systems.However, paging on broadcast disks differs from virtual memory paging in at least two fundamental aspects: l A page fault in the broadcast disk model has a variable cost that depends on the requested page as well as the current state of the broadcast.l Prefetching is both natural and a provably essential mechanism for achieving significantly better competitive ratios in broadcast disk paging.In this paper, we design a deterministic algorithm that uses prefetching to achieve an O(nlog k) competitive ratio for the broadcast disk paging problem, where k denotes the sire of the client's cache.We also show a matching lower bound of s2(nlog k) that applies even when the adversary is not allowed to use prefetching.In contrast, we show that when prefetching is not allowed, no deterministic online algorithm can achieve a competitive ratio better than D(nh).Moreover, we show a lower bound of SI(nlog k) on the competitive ratio achievable by any non-prefetching randomized algorithm against an oblivious adversary.These lower bounds are trivially matched from above by known results about deterministic and randomized marking algorithms for paging.An interpretation of our results is that in *Deut. of %t&mental