In　this　paper,　we　theoretically　analyse　the　uplink　(UL)　and　downlink　(DL)　performance　of　massive　multiple-input　and　multiple-output　(mMIMO)　networks,　in　term　of　coverage　probability,　cell　spectral　efficiency　and　network　area　spectral　efficiency,　using　stochastic　geometry.　A　sophisticated　but　yet　practical　system　model　is　considered,　taking　into　account　a　path　loss　model　differentiating　line-of-sight　and　non-line-of-sight　transmissions,　an　idle　mode　capability　at　the　base　stations　and　a　finite　user　density.　Our　analysis　pays　particular　attention　to　the　existence　of　a　finite　number　of　UL　pilots　for　channel　estimation　and　the　effect　of　pilot　contamination.　We　study　for　the　first　time　the　joint　impact　of　the　number　of　UL　pilot　sequences,　the　user　density,　and　the　base-station　density　on　the　pilot　contamination　issue　in　a　mMIMO　network,　which　in　turn　characterizes　the　DL　and　UL　network　performance.　Moreover,　using　the　proposed　framework,　we　investigate　two　different　scheduling　problems　-　UE　and　pilot　scheduling　-　,　to　find　the　optimal　simultaneously　scheduled　UE　density　per　time-frequency　resource　as　well　as　the　optimal　UL　pilot　number　to　maximise　the　spectral　efficiency.　©　2002-2012　IEEE.