Going mobile [mobile-computing networks]

Discusses virtual versus physical process-migration strategies in modeling mobile-computing networks. Both VPM and PPM play useful and effective roles in the modeling and simulation of real-world, mobile-computing networks. While PPM is a more accurate model of reality, by its very nature, VPM helps realize the modeling of networks with large numbers of mobile entities on testbeds with a modest number of processors, although at greatly reduced performance. Utamaphethai and Ghosh [1998] reported a TVHS simulation with 45,000 entities representing autonomous vehicles on a network of 60+ SUN Sparc 10 workstations. In contrast, a PPM implementation would be difficult to realize due to the large number of processors required and it would be unacceptably slow due to the large numbers of connections and disconnections. PPM is unquestionably superior when the computational loads associated with the mobile entities are high. However, its performance suffers relative to VPM where the system demands intense data exchange between the mobile and stationary units. This study also underscores that, for scalability in performance, the algorithms underlying real-world mobile computing networks must be designed to reduce communication while increasing local computation.