The way computing is done today is changing as a result of the everincreasing processing, storage, and communication capacities of modern computer hardware. Resource sharing scenarios, in which a physical server is shared for different applications, are becoming much more common. These scenarios require special attention to guarantee that the performance isolation of each application is carried out exactly as if it were being locally executed. Based on this situation, the present work aims at investigating techniques for providing resource reservations and thus guaranteeing quality of service (QoS) and performance isolation for applications. Considering environments in which the use of operating system extensions or the use of virtualization are unreasonable or inappropriate, this work investigates the viability and effectiveness of reservations done at user level, that is, reservations guaranteed with no operating system kernel instrumentation. For this purpose, we have implemented a tool that limits and ensures the proper usage of processing and disk bandwidth resources through the exclusive use of Linux Operating System primitives, which, among other functions, permits easy scheduling policy extensions. This feature enables flexibility in how resources are shared among distinct processes. Through tool usage analysis, we have identified the advantages and limitations of the techniques used. For a case study, aiming to achieve some specific performance goals, we have established parameter reservations for a three-tier application. We were able to verify that, even for complex applications, simple methodologies like linear regressions are capable of predicting resource usage with a low margin of error.