Mapping Parallel Iterative Algorithms onto Workstation Networks

For communication-intensive parallel applications, the maximum degree of concurrency achievable is limited by the communication throughput made available by the network. In previous work [HPS94], we showed experimentally that the performance of certain parallel applications running on a workstation network can be improved significantly if a congestion control protocol is used to enhance network performance. In this paper, we characterize and analyze the communication requirements of a large class of supercomputing applications that fall under the category of fixed-point problems, amenable to solution by parallel iterative methods. This results in a set of interface and architectural features sufficient for the efficient implementation of the applications over a large-scale distributed system. In particular, we propose a direct link between the application and network layer, supporting congestion control actions at both ends. This in turn enhances the system's responsiveness to network congestion, improving performance. Measurements are given showing the efficacy of our scheme to support large-scale parallel computations.

Measuring Bottleneck Link Speed in Package-Switched Networks

The quality of available network connections can often have a large impact on the performance of distributed applications. For example, document transfer applications such as FTP, Gopher and the World Wide Web suffer increased response times as a result of network congestion. For these applications, the document transfer time is directly related to the available bandwidth of the connection. Available bandwidth depends on two things: 1) the underlying capacity of the path from client to server, which is limited by the bottleneck link; and 2) the amount of other traffic competing for links on the path. If measurements of these quantities were available to the application, the current utilization of connections could be calculated. Network utilization could then be used as a basis for selection from a set of alternative connections or servers, thus providing reduced response time. Such a dynamic server selection scheme would be especially important in a mobile computing environment in which the set of available servers is frequently changing. In order to provide these measurements at the application level, we introduce two tools: bprobe, which provides an estimate of the uncongested bandwidth of a path; and cprobe, which gives an estimate of the current congestion along a path. These two measures may be used in combination to provide the application with an estimate of available bandwidth between server and client thereby enabling application-level congestion avoidance. In this paper we discuss the design and implementation of our probe tools, specifically illustrating the techniques used to achieve accuracy and robustness. We present validation studies for both tools which demonstrate their reliability in the face of actual Internet conditions; and we give results of a survey of available bandwidth to a random set of WWW servers as a sample application of our probe technique. We conclude with descriptions of other applications of our measurement tools, several of which are currently under development.

Performance Analysis of a WWW Server

This paper addresses the problem of analyzing performance of WWW servers. The web has experienced a phenomenal growth and has become the most popular Internet application. As a consequence of its large popularity, the Internet has suffered from various performance problems, such as network congestion and overloaded servers. These days, it is not uncommon to find servers refusing connections because they are overloaded. Performance has always been a key issue in the design and operation of on-line systems. With regard to Internet, performance is also critical, because users want fast and easy access to all objects (i.e., documents, pictures, audio, and video) available on the net. Thus, it is important to understand WWW performance issues. This paper focuses on the performance analysis of a Web server. Using a synthetic benchmark (WebStone), we analyze three different Web server software running on top of a Windows NT platform and performing some typical WWW tasks.

Query cost estimation through remote system contention states analysis over the Internet

Query processing over the Internet involving autonomous data sources is a major task in data integration. It requires the estimated costs of possible queries in order to select the best one that has the minimum cost. In this context, the cost of a query is affected by three factors: network congestion, server contention state, and complexity of the query. In this paper, we study the effects of both the network congestion and server contention state on the cost of a query. We refer to these two factors together as system contention states. We present a new approach to determining the system contention states by clustering the costs of a sample query. For each system contention state, we construct two cost formulas for unary and join queries respectively using the multiple regression process. When a new query is submitted, its system contention state is estimated first using either the time slides method or the statistical method. The cost of the query is then calculated using the corresponding cost formulas. The estimated cost of the query is further adjusted to improve its accuracy. Our experiments show that our methods can produce quite accurate cost estimates of the submitted queries to remote data sources over the Internet.

Control design for multicast-aware class-based networks

The expectations of citizens from the Information Technologies (ITs) are increasing as the ITs have become integral part of our society, serving all kinds of activities whether professional, leisure, safety-critical applications or business. Hence, the limitations of the traditional network designs to provide innovative and enhanced services and applications motivated a consensus to integrate all services over packet switching infrastructures, using the Internet Protocol, so as to leverage flexible control and economical benefits in the Next Generation Networks (NGNs). However, the Internet is not capable of treating services differently while each service has its own requirements (e.g., Quality of Service - QoS). Therefore, the need for more evolved forms of communications has driven to radical changes of architectural and layering designs which demand appropriate solutions for service admission and network resources control. This Thesis addresses QoS and network control issues, aiming to improve overall control performance in current and future networks which classify services into classes. The Thesis is divided into three parts. In the first part, we propose two resource over-reservation algorithms, a Class-based bandwidth Over-Reservation (COR) and an Enhanced COR (ECOR). The over-reservation means reserving more bandwidth than a Class of Service (CoS) needs, so the QoS reservation signalling rate is reduced. COR and ECOR allow for dynamically defining over-reservation parameters for CoSs based on network interfaces resource conditions; they aim to reduce QoS signalling and related overhead without incurring CoS starvation or waste of bandwidth. ECOR differs from COR by allowing for optimizing control overhead minimization. Further, we propose a centralized control mechanism called Advanced Centralization Architecture (ACA), that uses a single state-full Control Decision Point (CDP) which maintains a good view of its underlying network topology and the related links resource statistics on real-time basis to control the overall network. It is very important to mention that, in this Thesis, we use multicast trees as the basis for session transport, not only for group communication purposes, but mainly to pin packets of a session mapped to a tree to follow the desired tree. Our simulation results prove a drastic reduction of QoS control signalling and the related overhead without QoS violation or waste of resources. Besides, we provide a generic-purpose analytical model to assess the impact of various parameters (e.g., link capacity, session dynamics, etc.) that generally challenge resource overprovisioning control. In the second part of this Thesis, we propose a decentralization control mechanism called Advanced Class-based resource OverpRovisioning (ACOR), that aims to achieve better scalability than the ACA approach. ACOR enables multiple CDPs, distributed at network edge, to cooperate and exchange appropriate control data (e.g., trees and bandwidth usage information) such that each CDP is able to maintain a good knowledge of the network topology and the related links resource statistics on real-time basis. From scalability perspective, ACOR cooperation is selective, meaning that control information is exchanged dynamically among only the CDPs which are concerned (correlated). Moreover, the synchronization is carried out through our proposed concept of Virtual Over-Provisioned Resource (VOPR), which is a share of over-reservations of each interface to each tree that uses the interface. Thus, each CDP can process several session requests over a tree without requiring synchronization between the correlated CDPs as long as the VOPR of the tree is not exhausted. Analytical and simulation results demonstrate that aggregate over-reservation control in decentralized scenarios keep low signalling without QoS violations or waste of resources. We also introduced a control signalling protocol called ACOR Protocol (ACOR-P) to support the centralization and decentralization designs in this Thesis. Further, we propose an Extended ACOR (E-ACOR) which aggregates the VOPR of all trees that originate at the same CDP, and more session requests can be processed without synchronization when compared with ACOR. In addition, E-ACOR introduces a mechanism to efficiently track network congestion information to prevent unnecessary synchronization during congestion time when VOPRs would exhaust upon every session request. The performance evaluation through analytical and simulation results proves the superiority of E-ACOR in minimizing overall control signalling overhead while keeping all advantages of ACOR, that is, without incurring QoS violations or waste of resources. The last part of this Thesis includes the Survivable ACOR (SACOR) proposal to support stable operations of the QoS and network control mechanisms in case of failures and recoveries (e.g., of links and nodes). The performance results show flexible survivability characterized by fast convergence time and differentiation of traffic re-routing under efficient resource utilization i.e. without wasting bandwidth. In summary, the QoS and architectural control mechanisms proposed in this Thesis provide efficient and scalable support for network control key sub-systems (e.g., QoS and resource control, traffic engineering, multicasting, etc.), and thus allow for optimizing network overall control performance.

Estudo e proposta de algoritmo para a minimização do corte de eólica no contexto de uma análise custo-benefício multicritério

Trabalho Final de Mestrado para obtenção do grau de Mestre em Engenharia Electrotécnica Ramo de Energia

A genetic algorithm for the design of a fuzzy controller for active queue management

Active queue management (AQM) policies are those policies of router queue management that allow for the detection of network congestion, the notification of such occurrences to the hosts on the network borders, and the adoption of a suitable control policy. This paper proposes the adoption of a fuzzy proportional integral (FPI) controller as an active queue manager for Internet routers. The analytical design of the proposed FPI controller is carried out in analogy with a proportional integral (PI) controller, which recently has been proposed for AQM. A genetic algorithm is proposed for tuning of the FPI controller parameters with respect to optimal disturbance rejection. In the paper the FPI controller design metodology is described and the results of the comparison with random early detection (RED), tail drop, and PI controller are presented.

Parallel genetic algorithms for the tuning of a fuzzy AQM controller

This paper presents the results of the application of a parallel Genetic Algorithm (GA) in order to design a Fuzzy Proportional Integral (FPI) controller for active queue management on Internet routers. The Active Queue Management (AQM) policies are those policies of router queue management that allow the detection of network congestion, the notification of such occurrences to the hosts on the network borders, and the adoption of a suitable control policy. Two different parallel implementations of the genetic algorithm are adopted to determine an optimal configuration of the FPI controller parameters. Finally, the results of several experiments carried out on a forty nodes cluster of workstations are presented.

Protect grids from DDoS attacks

Recently a number of highly publicised incidents of Distributed Denial of Service (DDoS) attacks have made people aware of the importance of providing available securely the grids’ data and services to users. This paper introduces the vulnerability of grids to DDoS attacks, and proposes a distributed defense system that has a mixture deployment of sub-systems to protect grids from DDoS attacks. According to the simulation experiments, this system is effective to defend grids against attacks. It can avoid overall network congestion and provide more resources to legitimate grid users.

Protecting web applications from DDoS attacks by an active distributed defense system

In the last a few years a number of highly publicized incidents of Distributed Denial of Service (DDoS) attacks against high-profile government and commercial websites have made people aware of the importance of providing data and services security to users. A DDoS attack is an availability attack, which is characterized by an explicit attempt from an attacker to prevent legitimate users of a service from using the desired resources. This paper introduces the vulnerability of web applications to DDoS attacks, and presents an active distributed defense system that has a deployment mixture of sub-systems to protect web applications from DDoS attacks. According to the simulation experiments, this system is effective in that it is able to defend web applications against attacks. It can avoid overall network congestion and provide more resources to legitimate web users.

Load balance algorithms for anycast

Increasingly, replicated anycast servers are being used to deliver network applications and service ever increasing user requests. Therefore, the strategies used to guarantee network bandwidth prerequisites and perform load balancing across the nodes of an anycast group are critical to the performance of online applications. In this paper, we model user requests, network congestion and latency, and server load using a combination of hydro-dynamics and queuing theory to develop an efficient job distribution strategy. Current, anycast research does not explicitly consider the system load of nodes within an anycast groups when distributing requests. Therefore, the performance of a heavily loaded anycast system can quickly become congested and uneven as jobs are routed to closely linked nodes which are already saturated with requests. In comparison, the nodes of further away systems remain relatively unused because of other issues such as network bandwidth and latency during these times. Our system redirects requests from busy systems to the idle, remotely linked nodes, to process requests faster in spite of slower network access. Using an empirical study, we show this technique can improve request performance, and throughput with minimal network probing overhead.

A Comparative Analysis of Beaconless Opportunistic Routing Protocols for Video Dissemination over Flying Ad-Hoc Networks

A reliable and robust routing service for Flying Ad-Hoc Networks (FANETs) must be able to adapt to topology changes, and also to recover the quality level of the delivered multiple video flows under dynamic network topologies. The user experience on watching live videos must also be satisfactory even in scenarios with network congestion, buffer overflow, and packet loss ratio, as experienced in many FANET multimedia applications. In this paper, we perform a comparative simulation study to assess the robustness, reliability, and quality level of videos transmitted via well-known beaconless opportunistic routing protocols. Simulation results shows that our developed protocol XLinGO achieves multimedia dissemination with Quality of Experience (QoE) support and robustness in a multi-hop, multi-flow, and mobile networks, as required in many multimedia FANET scenarios.

Orquestração de migração massiva de máquinas virtuais baseada em análise cíclica para ambientes de computação na nuvem.

Um das principais características da tecnologia de virtualização é a Live Migration, que permite que máquinas virtuais sejam movimentadas entre máquinas físicas sem a interrupção da execução. Esta característica habilita a implementação de políticas mais sofisticadas dentro de um ambiente de computação na nuvem, como a otimização de uso de energia elétrica e recursos computacionais. Entretanto, a Live Migration pode impor severa degradação de desempenho nas aplicações das máquinas virtuais e causar diversos impactos na infraestrutura dos provedores de serviço, como congestionamento de rede e máquinas virtuais co-existentes nas máquinas físicas. Diferente de diversos estudos, este estudo considera a carga de trabalho da máquina virtual um importante fator e argumenta que escolhendo o momento adequado para a migração da máquina virtual pode-se reduzir as penalidades impostas pela Live Migration. Este trabalho introduz a Application-aware Live Migration (ALMA), que intercepta as submissões de Live Migration e, baseado na carga de trabalho da aplicação, adia a migração para um momento mais favorável. Os experimentos conduzidos neste trabalho mostraram que a arquitetura reduziu em até 74% o tempo das migrações para os experimentos com benchmarks e em até 67% os experimentos com carga de trabalho real. A transferência de dados causada pela Live Migration foi reduzida em até 62%. Além disso, o presente introduz um modelo que faz a predição do custo da Live Migration para a carga de trabalho e também um algoritmo de migração que não é sensível à utilização de memória da máquina virtual.

Predicting delay reductions from freeway traffic diversion for incident management

Traffic incidents are a major source of traffic congestion on freeways. Freeway traffic diversion using pre-planned alternate routes has been used as a strategy to reduce traffic delays due to major traffic incidents. However, it is not always beneficial to divert traffic when an incident occurs. Route diversion may adversely impact traffic on the alternate routes and may not result in an overall benefit. This dissertation research attempts to apply Artificial Neural Network (ANN) and Support Vector Regression (SVR) techniques to predict the percent of delay reduction from route diversion to help determine whether traffic should be diverted under given conditions. The DYNASMART-P mesoscopic traffic simulation model was applied to generate simulated data that were used to develop the ANN and SVR models. A sample network that comes with the DYNASMART-P package was used as the base simulation network. A combination of different levels of incident duration, capacity lost, percent of drivers diverted, VMS (variable message sign) messaging duration, and network congestion was simulated to represent different incident scenarios. The resulting percent of delay reduction, average speed, and queue length from each scenario were extracted from the simulation output. The ANN and SVR models were then calibrated for percent of delay reduction as a function of all of the simulated input and output variables. The results show that both the calibrated ANN and SVR models, when applied to the same location used to generate the calibration data, were able to predict delay reduction with a relatively high accuracy in terms of mean square error (MSE) and regression correlation. It was also found that the performance of the ANN model was superior to that of the SVR model. Likewise, when the models were applied to a new location, only the ANN model could produce comparatively good delay reduction predictions under high network congestion level.

Predicting delay reductions from freeway traffic diversion for incident management

Traffic incidents are a major source of traffic congestion on freeways. Freeway traffic diversion using pre-planned alternate routes has been used as a strategy to reduce traffic delays due to major traffic incidents. However, it is not always beneficial to divert traffic when an incident occurs. Route diversion may adversely impact traffic on the alternate routes and may not result in an overall benefit. This dissertation research attempts to apply Artificial Neural Network (ANN) and Support Vector Regression (SVR) techniques to predict the percent of delay reduction from route diversion to help determine whether traffic should be diverted under given conditions. The DYNASMART-P mesoscopic traffic simulation model was applied to generate simulated data that were used to develop the ANN and SVR models. A sample network that comes with the DYNASMART-P package was used as the base simulation network. A combination of different levels of incident duration, capacity lost, percent of drivers diverted, VMS (variable message sign) messaging duration, and network congestion was simulated to represent different incident scenarios. The resulting percent of delay reduction, average speed, and queue length from each scenario were extracted from the simulation output. The ANN and SVR models were then calibrated for percent of delay reduction as a function of all of the simulated input and output variables. The results show that both the calibrated ANN and SVR models, when applied to the same location used to generate the calibration data, were able to predict delay reduction with a relatively high accuracy in terms of mean square error (MSE) and regression correlation. It was also found that the performance of the ANN model was superior to that of the SVR model. Likewise, when the models were applied to a new location, only the ANN model could produce comparatively good delay reduction predictions under high network congestion level.