Rapid Simulation of Wireless Systems L. Felipe Perrone Dept. of Computer Science College of William and Mary perron,e@cs.‘wm. edu Abstract The sim.ulation of wireless systems such as cellular o’r personal communication systems comprises both dis- crete and continuous tim,e processes. To accelerate these simulatio*ns, we propose the use of interval jump- ing, a novel technique that allows the execution of a continuous time model to proceed in irregularly sized jumps rather than in the traditional time-stepped ma’n- ner. The foundations for this mechanism are laid out in the light of the simvlation of a complex simu,la,tion model which includes radio propagation, channel alloca- tion, transmitter power control a,nd user mobility. We conclude with cxperimen,tal results comparing sequen- tial and parallel execution of these accelerated simula- tions which indicate the good potential of our technique. 1. An Overview of Wireless Systems Cellular phone systems, paging networks and per- sonal communication systems (PCS) are particular in- stances of the broader class of wireless systems. A central idea common to these is that the spatial do- main is divided into smaller areas of service called cells. Within each cell, mobile subscribers use low power ra- dio transceivers to connect to a server positioned at its cerner. The servers route user calls within the wireless system and/or to a conventional wired network. The radio link used can be analog, digital or hybrid, supporting both types of signals. In any case, some means must be used to provide server access to multiple subscribers at the same time. Depending on the nature of the radio signal, different technologies are available, such as Time Division Multiple Access (TDMA), Fre quency Division Multiple Access (FDMA), etc. Regardless of the implementation of multiple ac- cess, radio connections must use the lowest possible transmission power. The same channel may be used by a number of cells simultaneously and low power helps minimize cocha,nnel interference allowing a cer- tain quality of se~rvicc (QoS) to be achieved. This is the concept of chaaael reuse, one of the central ideas behind the development of wireless systems. David M. Nicol Dept. of Computer Science Dartmouth College nicol@cs. dartmouth. edu The QoS of a system is defined in terms of the ratio of strength of the signal generated by a transmitter measured at the receiver’s geographic location to the total noise perceived at the same point. This ratio is commonly referred to as signal-to-noise ratio (SNR). Service providers determine that a connection is only as good as its SNR: it is preferable to not offer or interrupt a connection if the clarity of signal does not achieve a desired standard. Other measures of importance in wireless systems are blocking probability, the likelihood that a call attempt fails due to system saturation, and drop probability, the likelihood that a call is interrupted. A call, may be blocked either if the cell that provides service has no more free channels or has no free channel that can per- form at the desired SNR. A call may be dropped when the user moves outside the current service area into another that has no free channels or if no channel can carry it at the target SNR. We can say that the three values, SNR, blocking and drop probabilities, define an overall performance measure of a wireless system. It is difficult to predict the performance of a specific wireless system because there are no analytical models that embody the complicated dynamics of all its com- ponents, what justifies the use of computer simulation. The models for these systems, however, are usually so complex and big that simulation times can easily reach an order of tens of hours. In this work we identify a major bottleneck in these simulations and propose a new technique to overcome it. Section 2 presents a brief discussion of related work to situate this effort within the literature. Section 3 defines our simulation model and basic terminology. Section 4 introduces interval jumping and sections 5 and 6 present experimental results with sequential and parallel simulations, respectively. 2. Related Work The latest years have witnessed the publication of several articles on parallel simulation of wireless sys- tems. The level of abstraction in the models used in these works varies according with the depth and com- 170 1087-4097/98 $10.00 0 1998 IEEE