This paper investigates the problem of server parameter selection in hierarchical fixed priority preemptive systems. A set of algorithms are provided that determine the optimal values for a single server parameter (capacity, period, or priority) when the other two parameters are fixed. By contrast, the general problem of server parameter selection is shown to be a holistic one: typically the locally optimal solution for a single server does not form part of the globally optimal solution. Empirical investigations show that improvements in remaining utilisation (spare capacity) can be achieved by choosing server periods that are exact divisors of their task periods; enabling tasks to be bound to the release of their server, enhancing task schedulability and reducing server capacity requirements.