Abstract : During the last few years, the rise of the Internet has changed the way business is conducted worldwide. To remain competitive, businesses have been implementing information technology support for business processes over the years. In this context, Service Oriented Architectures (SOA) have emerged as the main solution for the integration of legacy systems with new technologies within large organizations. Modern Enterprise Data Centers (EDCs) implementing SOA concepts and solutions are usually deployed as a two-tiered architecture where, in order to relieve service servers from the computational cost of CPU intensive tasks (e.g., XML parsing) and to perform resource protection, these functions are offloaded on a cluster of SON (Service-Oriented Networking) appliances. In EDC setups, access to services is governed by Service-Level Agreements (SLAs), which aim at protecting EDC resources. Currently, SON appliances are able to protect EDC resources by limiting the access (i.e., controlling the traffic) to services. Resource provisioning and optimization is a classic QoS management problem. Moreover, traffic control is a well-known problem in network traffic engineering. However, in service-oriented EDC setups the problem is fundamentally different. In classic networks, the resource protected by the shaping function is typically link bandwidth and buffer space, the units of which are precisely defined and measurable. In an EDC environment, resource metrics mostly fall into one of the following types : CPU power and main memory from application servers (CPU and memory), disk storage from storage servers (disk), and link bandwidth on the internal EDC network (bandwidth). Another fundamental difference is that in ''classic'' networks traffic control has local scope, since traffic is in the form of a single connection. In an EDC environment, service clients access services from multiple entry points (e.g., a cluster of SON appliances). Thus, the desired effect is ''global'' shaping. The challenge is then to enforce contracts by taking local actions at each entry point. The contributions of these thesis are threefold. We first propose and validate DoWSS, a dynamic credit-based algorithm for multipoint-to-point service traffic shaping. Contrary to existing credit-based approaches, DoWSS involves the use of a doubly-weighted strategy for credit allocation. The evaluation results of DoWSS show that it performs optimally by limiting the number of requests to maximum possible number allowed by the client service contract. Second, we argue that current off-the-shelf SON appliances present architectural limitations that prevent them from being used to efficiently perform traffic shaping in the presence of multiple service hosts. To tackle this issue, we introduce MuST, a SON Appliance architecture fit for multi-service traffic shaping. Our validation via simulation shows that our approach solves the multipoint-to-multipoint service traffic shaping problem while pushing the system to its maximum capacity. Finally, current trends point to having applications located in geographically distributed EDCs. Existing traffic shaping approaches, which are designed for single-site EDCs, present issues related to network latencies when used in geographically distributed environments. To tackle this issue, we propose GeoDS, a geographically distributed service traffic shaping approach that considers in its design the communications delays between entities in the system. Our evaluation shows that our approach is able to efficiently solve the service traffic shaping problem in geographically distributed environments.