Recently, increasingly large amounts of data are generatedfrom a variety of sources. Existing data processing technologiesare not suitable to cope with the huge amountsof generated data. Yet, many research works focus on BigData, a buzzword referring to the processing of massive volumesof (unstructured) data. Recently proposed frameworksfor Big Data applications help to store, analyze and processthe data. In this paper, we discuss the challenges of BigData and we survey existing Big Data frameworks. We alsopresent an experimental evaluation and a comparative studyof the most popular Big Data frameworks with several representativebatch and iterative workloads. This survey isconcluded with a presentation of best practices related tothe use of studied frameworks in several application domainssuch as machine learning, graph processing and real-worldapplications.