As the Internet and Telecommunications domains merge, open telecommunication service architectures such as TINA, PARLAY and PINT are becoming prevalent. Distributed Computing is a common engineering component in these technologies and promises to bring improvements to the scalability, reliability and flexibility of telecommunications service delivery systems. This distributed approach to service delivery introduces new performance concerns. As service logic is decomposed into software components and distnbuted across network resources, significant additional resource loading is incurred due to inter-node communications. This fact makes the choice of distribution of components in the network and the distribution of load between these components critical design and operational issues which must be resolved to guarantee a high level of service for the customer and a profitable network for the service operator. Previous research in the computer science domain has addressed optimal placement of components from the perspectives of minimising run time, minimising communications costs or balancing of load between network resources. This thesis proposes a more extensive optimisation model, which we argue, is more useful for addressing concerns pertinent to the telecommunications domain. The model focuses on providing optimal throughput and profitability of network resources and on overload protection whilst allowing flexibility in terms of the cost of installation of component copies and differentiation in the treatment of service types, in terms of fairness to the customer and profitability to the operator. Both static (design-time) component distribution and dynamic (run-time) load distribution algorithms are developed using Linear and Mixed Integer Programming techniques. An efficient, but sub-optimal, run-time solution, employing Market-based control, is also proposed. The performance of these algorithms is investigated using a simulation model of a distributed service platform, which is based on TINA service components interacting with the Intelligent Network through gateways. Simulation results are verified using Layered Queuing Network analytic modelling Results show significant performance gains over simpler methods of performance control and demonstrate how trade-offs in network profitability, fairness and network cost are possible.