The　increasing　maturity　of　Cloud　computing　technology　brings　a　profound　change　to　modern　information　service　industry.　With　explosive　growth　of　the　web　services,　traditional　service　registration　software　encounters　many　problems　on　cloud　service　registration　and　discovery.　In　this　paper,　we　design　and　implement　a　novel　geographical　knowledge　service　registry　(GeoSC)　to　address　the　problems　of　organizing　and　managing　vast　amounts　of　heterogeneous　knowledge　services　under　cloud　environment　and　support　for　fast,　accurate　service　discovery.　Based　on　open-source　cloud　computing　framework　-　Hadoop　stack,　GeoSC　adopted　a　non-relational　data-storage　model　to　support　QoS-based　and　semantics-based　service　description,　and　provide　efficient　parallel　service　discovery.　We　carried　out　an　elaborate　comparison　between　GeoSC　and　traditional　service　registration　software　-　JUDDI　to　validate　web　service　registration　and　discovery　performance　improvement.　The　preliminary　results　indicated　that,　GeoSC　performs　better　on　service　registration　and　discovery　than　JUDDI.　©　2013　IEEE.