From the perspective of performance and parts cost, distributed architectures are often superior to single processor architectures for embedded systems designs. Despite this, embedded system designers still tend to use single processors whenever possible. One of the main reasons for this is that the additional costs in design and maintenance of distributed systems can far outweigh the lower parts cost and lower power consumption that might result from a distributed implementation. This is largely because most distributed models are limited in the system level abstraction that can be derived.