We have developed an ad-hoc wireless positioning network (AWPN) to realize on-demand indoor location-based services . This paper extends our AWPN to handle huge number of localization requests. In AWPN, WiFi APs measure received signal strength (RSS) of WiFi signals and send the RSS information to a localization server via a WiFi mesh network. The maximum number of WiFi devices is therefore limited by computational resources on the localization server. We push this limit by introducing a new distributed calculation scheme: we use the MapReduce computation framework and perform map processes on APs and reduce processes on localization servers. We also utilize a network router capable of network address translation (NAT) for shuffle processes to provide scalability. We implemented and evaluated our distributed calculation scheme to demonstrate that our scheme almost evenly distributes localization calculations to multiple localization servers with approximately 26% variations.