Abstract : Big data stream analytics platforms not only need to support performance-dictated elasticity benefiting for instance from Cloud environments. They should also support analytics that can evolve dynamically from the application viewpoint, given data nature can change so the necessary treatments on them. The benefit is that this can avoid to undeploy the current analytics, modify it off-line, redeploy the new version, and resume the analysis, missing data that arrived in the meantime. We also believe that such evolution should better be driven by autonomic behaviors whenever possible. We argue that a software component based technology, as the one we have developed so far, GCM/ProActive, can be a good fit to these needs. Using it, we present our solution, still under development, named GCM-streaming, which to our knowledge seems to be quite original.