We consider a multi-processor system-on-chip destined for streaming applications. An application is composed of one input and one output queue and in-between, several levels of identical tasks. Data arriving at the input are treated in parallel in an arbitrary order, but have to leave the system in the order of arrival. This scenario is particularly important in the context of telecommunication applications, where the duration of treatment depends on the packets’ contents. We present an algorithm which re-establishes the packet order: packets are dropped if their earliness or lateness exceeds a limit previously fixed by experimentation; otherwise, they are stored in a buffer on the output side. Write operations to this buffer are random access, whereas read operations are in FIFO order. Our algorithm guarantees that no data is removed from the queue before it has been read. For a given throughput, we guarantee a minimum buffer size. We implemented our algorithm within the output coprocessor in the form of communicating finite state machines and validated it on a multi-processor telecommunication platform.