FundaÃÃo de Amparo à Pesquisa do Estado do Cearà / Web Services (WS) sÃo pilares para a construÃÃo de aplicaÃÃes orientadas a serviÃos. Uma sÃrie de linguagens para a composiÃÃo de serviÃos web tÃm sido propostas, sendo formado um consenso em torno da linguagem de execuÃÃo de processos de negÃcio (BPEL). BPEL centra-se em processos de negÃcio que orquestram interaÃÃes de WS. No entanto, em geral, processos de negÃcio sÃo compostos por um amplo espectro de atividades que exigem muitas vezes a participaÃÃo humana para executar tarefas, rever ou aprovar medidas e inserir dados. Essas interaÃÃes humanas sÃo abordadas em uma nova especificaÃÃo do BPEL denominada de BPEL4People. BPEL4People introduz a atividade humana para BPEL. Com o uso do BPEL4People, modelos formais (como as redes de Petri) de BPEL4People tÃm sido propostos. Com base em modelos formais à possÃvel a realizaÃÃo de anÃlises formais, tais como anÃlise de desempenho de modelos para descobrir possÃveis problemas em WS orquestrados com o BPEL4People. Hà duas dimensÃes importantes para o desempenho de WS: tempo de resposta e escalabilidade. O tempo de resposta à a capacidade de um sistema de prover tempos aceitÃveis para suas atividades e a escalabilidade à a capacidade de um sistema de continuar a cumprir seus objetivos de tempo de resposta quando a demanda pelo mesmo aumenta. Alguns trabalhos tÃm sido propostos acerca da anÃlise do desempenho de WS orquestrados com o BPEL. Nesta tese, à proposta uma arquitetura denominada ``SOASPE' (SOA + SPE) para a transformaÃÃo de cÃdigos BPEL4People em redes de Petri estocÃsticas generalizadas - Generalized Stochastic Petri Nets (GSPN) e redes de Petri coloridas - Coloured Petri Net (CPN). AtravÃs dos modelos GSPN e CPN de BPEL4People à possÃvel avaliar o desempenho dos WS orquestrados com o BPEL4People atravÃs da comparaÃÃo dos seus tempos de resposta reais quando submetidos a uma quantidade de requisiÃÃes e dos tempos de resposta dos modelos GSPN e CPN dos WS gerados pela arquitetura ``SOASPE' quando realizadas simulaÃÃes com a mesma quantidade de requisiÃÃes. Durante a transformaÃÃo de cÃdigos BPEL4People, as redes de Petri sÃo usadas para modelar atividades BPEL e as atividades humanas. EntÃo, pela simulaÃÃo dos modelos gerados, problemas potenciais com o desempenho de WS orquestrados com o BPEL4People podem ser detectados. / Web Services (WS) are pillars for the construction of service-oriented applications. A number of languages for web service composition have been proposed, formed a consensus on the language of business process execution (BPEL). BPEL focuses on business processes that orchestrate WS interactions. However, in general, business processes are composed of a broad spectrum of activities that often require human involvement to perform tasks, review or approve steps and enter data. These human interactions are discussed in a new specification of BPEL called BPEL4People. BPEL4People introduces human activity to BPEL. With the use of BPEL4People, formal models (such as Petri nets) of BPEL4People have been proposed. Based on formal models is possible to perform formal analysis, such as performance analysis of models to discover potential problems with the WS orchestrated with BPEL4People. There are two dimensions important to the performance of WS: response time and scalability. The response time is the ability of a system to provide an acceptable time for their activities and scalability is the ability of a system to continue to fulfill its goals of response time when demand for it increases. Some works have been proposed about the performance analysis of WS orchestrated with BPEL. In this thesi, it is proposed an architecture called ``SOASPE' (SOA + SPE) for the transformation of the BPEL4People code in generalized stochastic Petri nets (GSPN) and colored Petri nets (CPN). Through the GSPN and CPN models of BPEL4People is possible to evaluate the performance of WS orchestrated with BPEL4People by comparing their real response times when subjected to a number of requests and response times of the GSPN and CPN models generated by the architecture ``SOASPE' when simulations with the same amount of requests. During the transformation of BPEL4People code, Petri nets are used to model BPEL activities and human activities. Then, by simulation of generated models, potential problems with the performance of WS orchestrated with BPEL4People can be detected.

In a typical workflow process, exceptions are the norm. Exceptions are defined as deviations from the normal sequence of activities and events. Exceptions can be divided into two broad categories: known exceptions (i.e., expected and predefined deviations) and unknown exceptions (i.e., unexpected and undefined deviations). Business Process Execution Language (BPEL) has become the de facto standard for executing business workflows with the use of web services. BPEL includes exception handling methods that are sufficient for known exception scenarios. Depending on the exception and the specifics of the exception handling tools, processes may either halt or move to completion. Instances of processes that are halted or left incomplete due to unhandled exceptions affect the performance of the workflow process, as they increase resource utilization and process completion time. However, designing efficient process handlers to avoid the issue of unhandled exceptions is not a simple task. This thesis provides a tool that handles unknown exceptions using provisions for exception handling with the involvement of human activities by using the BPEL4PEOPLE specification. BPEL4PEOPLE, an extension of BPEL, offers the ability to specify human activities within BPEL processes. The approach considered in this thesis involves humans in exception handling tools by providing an alternate sub process within a given business process. A prototype application has been developed implementing the tool that handles unknown exceptions. The prototype application monitors the progress of an automated workflow process and permits human involvement to reroute the course of a workflow process when an unknown exception occurs. The utility of the prototype and the tool using the Scenario Walkthrough and Inspection Methods (SWIMs) are demonstrated. We demonstrate the utility of the tool through loan application process scenarios, and offer a walkthrough of the system by using examples of instances with examples of known and unknown exceptions, as well as a claims analysis of process instances results.