Cluster-based databases have been traditionally used for providing fault tolerant storage solutions. However, the limited scalability of this kind of systems has motivated the creation of a variety of storage systems based on the cloud paradigm, which achieve high scalability and availability by giving up some functionalities such as strong consistency. While this is fairly acceptable for a wide range of applications, there are still many use cases which cannot take advantage of the cloud paradigm because they are unable to resign their transactional nature. The purpose of this paper is to review how these novel systems overcome typical scalability and availability issues and present a new architecture able to offer transactional support on the cloud. This hybrid architecture is based on well-known replication protocols used in classic transactional systems which are combined according to the application workload, thus providing different consistency levels depending on the needs of each application. Moreover, the system inherits the elasticity and dynamism properties of the cloud paradigm, which makes it able to offer high scalability and availability levels without sacrificing transactional support.