Figure 3 - uploaded by Magnos Martinello
Content may be subject to copyright.
Source publication
Cette thèse porte sur le développement d’une approche de modélisation pragmatique permettant aux concepteurs d’applications et systèmes mis en oeuvre sur le web d’évaluer la disponibilité du service fourni aux utilisateurs. Plusieurs sources d’indisponibilité du service sont prises en compte, en particulier i) les défaillances matérielles ou logici...
Contexts in source publication
Context 1
... protocol basically processes each incoming request in a similar way as tcp does, but it never sends back any reply to the client. ¿º¿ ÅÓÐÐÒÒ ××ÙÑÔØØÓÒ× Figure 3.1 shows a simple example of a clustered architecture composed of multiple server nodes with a dispatcher that distributes the incoming requests among the nodes. For this study, we assume that the arriving traffic is dispatched to the nodes according to a round robin strategy. ...
Context 2
... restoration times are assumed to be exponentially distributed with rate . Figure 3.2 shows the availability model describing the behavior of the cluster governed by server failures, detection, recovery and restoration processes. ...
Context 3
... first assumption, illustrated in Figure 3.3, is that the traffic arriving to the web cluster is modeled as a Poisson process with rate req/s (requests per second) and is independent of the failure process. Assuming that there are available servers in the cluster with the input traffic being distributed among the servers, then this system has independent Poisson arrival processes each one with rate ¼ . ...
Context 4
... ËËÒ××ØØÚÚØÝ ØÓ ××ÖÚÚÚÚ ÖÖØØ Figure 3.7 shows the effect of the service rate on Í Í , considering two values of (5 and 10 req/sec). MTTD is set to ¾ sec., while the other parameters are set to their nominal values of Table 3.3. ...
Context 5
... ÁÑÔÔÔØ ÓÓ ØÖÖÖÖ ÑÓÐ Preliminary analysis of the access logs from the web server of our laboratory suggests that the daily cycles can be divided into five phases during which we assume the arrival process being stationary. Figure 3.8 provides examples of MMPP traffic model, for which the first phase (from 1h to 6h) consists of the lowest period of the day, then the traffic increases (from 6h to 9h) and continues to increase from 9h to 13h. ...
Context 6
... have considered three traffic models referred to as MMPP1, MMPP2, MMPP3 illustrated in figure 3.8. The values of the request rates were chosen for comparison purposes with the basic Poisson traffic with average arrival rate of ¾¼ req/sec. ...
Similar publications
21 st , 2002* service level agreement, web performance tuning, software performance modeling, response time, queueing, optimization We present a nonlinear integer optimization model for determining the number of machines at each tier in a multi-tier server network. We utilize a result from an open queueing network model on the average response time...
Modern data centers use virtualization as a means to increase utilization of increasingly powerful multi-core servers. Applications often require only a fraction of the resources provided by modern hardware. Multiple concurrent workloads are therefore required to achieve adequate utilization levels. Current virtualization solutions allow hardware t...
One of the major problems in building large-scale enterprise systems is anticipating the performance of the eventual solution before it has been built. This problem is especially germane to modern Internet-based e-business applications, where failure to provide high performance and scalability can lead to application and business failure. The funda...
Internet traffic often exhibits a structure with rich high-order statistical properties like self-similarity and long-range dependency (LRD). This greatly complicates the problem of server performance modeling and optimization. Existing tools like queuing models in most cases only hold in mean value analysis under the assumption of simplified traff...
Application servers have been widely adopted as
distributed infrastructure (or middleware) for developing
distributed systems. Current approaches for performance
evaluation of application servers have mainly
concentrated on the adoption of measurement techniques.
This paper, however, focuses on the use of simulation
techniques and presents an...
Citations
... Quantoà infraestrutura de redes, dependências adicionais têm sido impostas devidoà natureza distribuída e virtualizada das aplicações. Há uma necessidade eminente de oferecer garantias das propriedades de dependabilidade, em particular disponibilidade, confiabilidade e desempenho no serviço de conectividade [Martinello 2005]. Para oferecer tais garantias,é preciso entender o processo de caracterização de falhas ...
... Quantoà infraestrutura de redes, dependências adicionais têm sido impostas devidoà natureza distribuída e virtualizada das aplicações. Há uma necessidade eminente de oferecer garantias das propriedades de dependabilidade, em particular disponibilidade, confiabilidade e desempenho no serviço de conectividade [Martinello 2005]. Para oferecer tais garantias,é preciso entender o processo de caracterização de falhas nas suas múltiplas dimensões: localização, causas, tipos/classes, frequência e duração. ...
... De acordo com [Martinello 2005], uma falha de serviçoé uma transição do estado em que o serviço está funcionando de forma correta/esperada para o estado em que não esteja. No caso de um serviço de conectividade entre entidades x e y, uma falha ocorre quando a comunicação entre x e y fica impossibilitada ou ocorre com qualidade inferior a um certo limite tolerável. ...
A Rede Ipê é fundamental para a comunidade científica brasileira por interconectar universidades e centros de pesquisa de todo o país. Este artigo analisa algumas características da Rede Ipê e explora o uso de técnicas de aprendizado de máquina para predição de falhas em serviços de conectividade usando dados públicos disponibilizados pela ferramenta ViaIpê. O problema é abordado como uma tarefa de classificação binária utilizando redes neurais recorrentes. Os resultados mostram que a dependabilidade do serviço de conectividade varia significativamente nos diferentes PoPs da Rede Ipê. Além disso, apesar da heterogeneidade deste serviço, os modelos de predição mostram-se promissores, apresentando boa acurácia e boa precisão em alguns cenários.
... The response time to a remote Cloud server depends on several factors, such as customers' geographical location in relation to the server, the available bandwidth of communication channels and network interfaces, the number of concurrent user connections to the server, the rate of requests, the hardware configuration of the server etc. [11]. ...
... where Gthe available user goodput, Githe available goodput of i node, Glink_ithe available link to the i node goodput, Vithe size of the resource V, provided by node i. For a comparative evaluation of the average response time to a request for centralized client-server and distributed Cloud architectures, initial data ( Table 2) is collected based on the analysis of the research [11][12][13][14]. Analyzing the graphical representation of the two types of architectures' behavior with the number of concurrent users makes it possible to conclude that the selected set of input data can achieve a significant reduction of response time in the case of a distributed Cloud architecture with 15 or more concurrent users that all interact with each other. ...
The service and business models of a Distributed Cloud solution based on peer-to-peer technology are presented in this article, in order to im-plement the proposed solution. Methods for organizing the interaction between the participants' nodes and nodes that are non-participants in the Distributed Cloud are proposed. Passive replication is used to improve service reliability. A competitive analysis of existing solutions within the scope of a decentralization approach for content sharing is conducted. Average response time to a request for a centralized client-server and distributed Cloud architecture is estimated.
... Indeed a significant body of work has focused on various aspects of web performance evaluation. Although many efforts have been dedicated to analyze the availability of web hosts using measurement-based techniques [38,39], less emphasis has been put on the modeling of web service availability taking into account the impact of server node failures and performance degradations (see [40] for a review of the state-of-the art). In the following, we summarize the multi-level modeling framework proposed in [14] in order to address this gap. ...
This paper summarizes the state of knowledge and ongoing research on methods
and techniques for resilience evaluation, taking into account the
resilience-scaling challenges and properties related to the ubiquitous
computerized systems. We mainly focus on quantitative evaluation approaches
and, in particular, on model-based evaluation techniques that are commonly used
to evaluate and compare, from the dependability point of view, different
architecture alternatives at the design stage. We outline some of the main
modeling techniques aiming at mastering the largeness of analytical
dependability models at the construction level. Actually, addressing the model
largeness problem is important with respect to the investigation of the
scalability of current techniques to meet the complexity challenges of
ubiquitous systems. Finally we present two case studies in which some of the
presented techniques are applied for modeling web services and General Packet
Radio Service (GPRS) mobile telephone networks, as prominent examples of large
and evolving systems.
... This equation is general and can be applied in a similar way to user level (Kaâniche, Kanoun, & Martinello, 2003;Martinello 2005). ...
... Assuming that each server is modeled by an M/M/1/b queue, closed form-expressions have been obtained in (Martinello 2005) for L(k), L(kλ), and L(D k ) considering both NCT and CT recovery strategies. The traffic received by the system is balanced by the dispatcher among the k operational servers. ...
When deployed as infrastructure components of real-time business processes, service computing applications we rely on for our daily activities elicit the proper addressing of performance and dependability issues. While recent developments in service-oriented architectures have come a long way in many aspects, ranging from semantics and ontologies to frameworks and design processes, performance and dependability remains a research demanding field.
Performance and Dependability in Service Computing: Concepts, Techniques and Research Directions highlights current technological trends and related research issues in dedicated chapters without restricting their scope. This book focuses on performance and dependability issues associated with service computing and these two complementary aspects, which include concerns of quality of service (QoS), real-time constraints, security, reliability and other important requirements when it comes to integrating services into real-world business processes and critical applications.
... • M(σ j (F i )) is the performability of the services involved in execution scenario j; This equation is general and can be applied in a similar way to user level (Kaâniche, Kanoun, & Martinello, 2003;Martinello 2005). ...
... π k and π Dk are the steady state probabilities associated to states k and D k of Figure 8. Assuming that each server is modeled by an M/M/1/b queue, closed form-expressions have been obtained in (Martinello 2005) for L(k), L(kλ), and L(D k ) considering both NCT and CT recovery strategies. The traffic received by the system is balanced by the dispatcher among the k operational servers. ...
... This is a mediation service for SIP messages in an VoIP telecommunication system. Although the reliability/availability for telecommunication systems has been studied [4], [5], [6], [8], [9], [10], [11], [13], most of the studies focus on the system availability/reliability without considering the impact of outages on customers, and very few of them take into account the application logic running on top of the system that affects the user-experienced service reliability. In this paper we develop methods to compute a commonly used service (un)reliability measure for telecommunication systems called defects per million (DPM), which counts the number of calls not served per million attempted calls [7], [12]. ...
Defects per million (DPM), defined as the number of calls out of a million dropped due to failures, is an important service (un)reliability measure for telecommunication systems. Most previous research derives the DPM from steady-state system availability model. In this paper, we develop a novel method for DPM computation which takes into consideration not only system availability, but also the impact of service application as well as the transient behavior of failure recovery. We illustrate this approach using a real system which is the IBM SIP SLEE cluster. Our method takes into account software/hardware failures, different stages of recovery, different phases of call flow, retry attempts and the interactions between call flow and failure/recovery behavior.
... Thus, the result of the main service may be invalidated by simple faults and errors in imported services. Dependability analysis has to focus on creating a systemwide dependability model of the component models and evaluating the impact of the faults in the individual components, including the identification of dependability bottlenecks and the sensitivity analysis of the overall system to the components' dependability characteristics [1]. Based on the observation that Web services-based workflows fit into the class of systems composed of multiple operational phases characterized by potentially different requirements and goals, the paper attempts to apply methodologies and tools for dependability analysis based on the paradigm of Multiple Phased Systems (MPS, [2], [6], [7]) to this emerging category of dependability critical systems. ...
As Web service-based system integration recently became the main- stream approach to create composite services, the dependability of such systems becomes more and more crucial. Therefore, extensions of the common service composition techniques are urgently needed in order to cover dependability as- pects and a core concept for the dependability estimation of the target compos- ite service. Since Web services-based workflows fit into the class of systems composed of multiple phases, this paper attempts to apply methodologies and tools for dependability analysis of Multiple Phased Systems (MPS) to this emerging category of dependability critical systems. The paper shows how this dependability analysis constitutes a very useful support to the service provider in choosing the most appropriate service alternatives to build up its own com- posite service.
... After a set of transformations (see [4] for all details ), we obtain equation (10) ...
... We need to compute the probability that the system with c servers has i requests at steady-state denoted p i (c). Assuming that the sequence of interarrival times is described by independent and identical exponential random variables of rate λ (a traditional M/M/c in which p i (c) is well-known [2] with ρ = λ cµ ), we obtain for service unavailability the following closed-form equation (see [4] for all details) ...
In this paper, we introduce a simple analytical modeling approach for computing service unavailability due to long response time, for infinite and finite single-server systems as well as for multi-server systems. Closed-form equations of system unavailability based on the conditional response time distributions are derived and sensitivity analyses are carried out to analyze the impact of long response time on service unavailability. The evaluation provides practical quantitative results that can help distributed system developers in design decisions
... The evaluation of the web service availability taking into account the modeling assumptions presented in section 3. is carried out adopting a performability approach [11]. The idea consists in combining the results obtained from two models: a performance model and an availability model. ...
Internet is often used for transaction based applications such as online banking, stock trading, among many others where the service outages are unacceptable. It is important for designers of such applications to analyze how hardware, software and performance related failures affect the quality of service delivered to the users. This paper presents analytical models for evaluating the service availability of web cluster architectures. A composite performance and availability modeling approach is defined considering various causes of service unavailability. In particular, web cluster systems are modeled taking into account: two recovery strategies (client transparent and non client-transparent). Sensitivity analysis results are presented to show their impact on the web service availability. The obtained results are aimed at providing useful guidelines to web designers.
... After a set of transformations (see [4] for all details ), we obtain equation (10) ...
... We need to compute the probability that the system with c servers has i requests at steady-state denoted p i (c). Assuming that the sequence of interarrival times is described by independent and identical exponential random variables of rate λ (a traditional M/M/c in which p i (c) is well-known [2] with ρ = λ cµ ), we obtain for service unavailability the following closed-form equation (see [4] for all details) ...
In this paper, we introduce a simple analytical modeling approach for computing service unavailability due to long response time, for infinite and finite single-server systems as well as for multi-server systems. Closed-form equations of system unavailability based on the conditional response time distributions are derived and sensitivity analyses are carried out to analyze the impact of long response time on service unavailability. The evaluation provides practical quantitative results that can help distributed system developers in design decisions.
