We propose decentralized processing and combining in the uplink for a distributed MIMO (D-MIMO) network by a subset method. A subset that consists of a partial number of access points (APs) that are connected to a CPU is defined for each UE. A subset that consists of only one AP turns a D-MIMO network into “small cell” network; a subset that consists of all APs connected to the CPU give the same performance as a fully centralized method. The paper also shows that the subset method provides a very scalable trade-off between complexity and performance. It is simpler to realize as the processing can be distributed and parallelized. In the studied simulation scenarios, we show that it is possible to reach 85% of the performance upper bound by including only 20% of total APs in the subset, and to reach 95% of the performance upper bound by including 40% of APs in the subset.