The GBIF portal contains a description of most of the global biodiversity data. It faces two problems, namely the data availability and a poor expressiveness of queries, mainly due to a growing number of users which keep expressing new needs. To tackle these problems, we envision a scalable and relatively low cost solution. With this in mind, we propose a non-invasive and decentralized architecture for processing GBIF queries over a cloud infrastructure. We define a dynamic strategy for data distribution and queries processing algorithms that fit the GBIF requirements. We demonstrate the feasibility and efficiency of our solution by a prototype implementation which allows for processing extra query types, up to now unsupported by the GBIF portal.