This paper proposes an autonomous resource allocation system (RAS) for smart neighborhood areas in presence of distributed energy resources and storage systems, with the purpose of grid demand reduction (GDR). Different from the past research on RAS, most of which are not broken down into resource allocation of individual appliances, do not address practical implementation of RAS with communication systems, and do not consider realistic case scenarios with network latency and communication link failure, this paper presents an improved appliance-level RAS developed in four operational modes with a designed bidding mechanism to exchange energy among neighborhood members through a common storage facility, a hierarchical cloud-based two-layered communication architecture founded on message queuing telemetry transport protocol to implement local and global messaging required for the proposed RAS, and realistic case scenarios by considering data from a real-world residential area and utilizing a virtual wide area network emulator to emulate characteristics of a real network in order to investigate the effects of network latency or communication link failure on the implemented RAS. From the results of diverse scenarios, it could be observed that the proposed system performs effectively to achieve GDR, even if the communication system fails partially in the smart community under test.