Hybrid delay/blocking wireless systems with greedy resource allocation

This paper presents an analysis of a hybrid delay /blocking system model with wide applicability in wireless networks. Real time voice traffic operates under a "blocked calls cleared" regime while non-real-time data is delayable. These traffic types compete for a fixed number of servers under two schemes: (i) first come first served (FCFS) and (ii) pre-emption (PE) of ongoing data calls by new voice calls. Each voice call uses one server, while data calls are served in a First-In First-Out (FIFO) queue with capacity equal to all servers not occupied by voice. Voice and data calls are modelled as Poisson arrivals with exponential holding times, and the resulting continuous-time Markov process is analysed using a stochastic simulation of the state space. The results presented demonstrate trade-offs between the performance of real-time voice and delayable data traffic, as well as the differences between this hybrid system and conventional systems based solely on blocking or delay. We show that the use of the FCFS scheme causes data calls to monopolise the system resources at the expense of voice call blocking. The PE scheme remedies this by providing priority to voice calls. However, in demand scenarios where data is the dominant traffic, voice calls can receive unrealistically good performance at the expense of increased delay for data calls. Overall, PE results in a more efficient use of the system resources, especially with high demand.