As　the　development　of　the　semantic　web,　RDF　data　set　has　grown　rapidly,　thus　causing　the　query　problem　of　massive　RDF.　Using　distributed　technique　to　complete　the　SPARQL　(Simple　Protocol　and　RDF　Query　Language)　Query　is　a　new　way　of　solving　the　large　amounts　of　RDF　query　problem.　At　present,　most　of　the　RDF　query　strategies　based　on　Hadoop　have　to　use　multiple　MapReduce　jobs　to　complete　the　task,　resulting　in　waste　of　time.　In　order　to　overcome　this　drawback,　MRQJ　(using　MapReduce　to　query　and　join)　algorithm　is　proposed　in　the　paper,　which　firstly　uses　a　greedy　strategy　to　generate　join　plan,　then　only　one　MapReduce　job　should　be　created　to　get　the　query　results　in　SPARQL　query　execution.　Finally,　a　contrast　experiment　on　the　LUBM　(Lehigh　University　Benchmark)　test　data　set　is　conducted,　the　results　of　which　show　that　MRQJ　method　has　a　great　advantage　in　the　case　that　the　query　is　more　complicated.