Recent Herschel findings on filaments in nearby low-mass star-forming clouds clearly points to a new paradigm to explain the formation of high density gas in turbulent clouds leading to the protostellar collapse. These filaments are the locations for core fragmentation at roughly the local Jeans mass. The formation of massive stars and of rich stellar clusters in this new paradigm is however not yet understood. Massive elongated/filamentary structures, referred as ridges, are massive filaments observed in regions of high-mass stars formation which may host the formation of massive stars. They have large average densities and show large velocity dispersion, and are roughly as cold as their low-mass counterparts. This may indicate that a larger effective Jeans mass in these ridges due to additional turbulent support could explain a core fragmentation extending up to higher stellar masses. The level of turbulent support in ridges is however difficult to measure due a high level of dynamics (flows, rotation, infall) which may not represent well the level of true support (isotropic) for Jeans fragmentation. More generally the structure and properties of ridges/massive filaments is not well known and requires dedicated studies.I will present our most recent results obtained with Herschel and the IRAM 30m towards the DR21 ridge in Cygnus X. Several massive protostars are actually observed in the DR21 ridge confirming it is the birth place of massive stars. I will show that the whole large scale region is compatible with a global collapse of a 15 pc cloud of several 10s of thousands of solar masses. The most recent IRAM 30m observations show that the ridge is made of several sub-filaments which are all more massive than their counterparts in low-mass star forming regions. I will discuss the implications of these results in the context of the origin of massive stars.