Distributed multiple-input multiple-output (MIMO) promises a dramatic capacity increase. While significant theoretical work has been done on distributed MIMO at the physical layer, how to translate the physical layer innovation into tangible benefits to real networks remains open. In particular, realizing multi-point to multi-point MIMO involves the following challenges: (i) how to accurately synchronize multiple APs in phase and time in order to successfully deliver precoded signals to the clients, and (ii) how to develop a MAC protocol to effectively support multi-point to multi-point MIMO. In this paper, we develop a practical approach to address the above challenges. We implement multi-point to multi-point MIMO for both uplink and downlink to enable multiple APs to simultaneously communicate with multiple clients. We examine a number of important MAC design issues, such as how to access the medium, perform rate adaptation, support acknowledgments in unicast traffic, deal with losses/collisions, and schedule transmissions. We demonstrate its feasibility and effectiveness through a prototype implementation on USRP and SORA, two of the most well-known software defined radio platforms.
