Streaming　data　processing　has　become　a　hot　topic　in　the　big　data　research.　To　ensure　the　timeliness　of　data　processing,　it　is　important　to　explore　the　performance-sensitive　components　in　the　streaming　data　processing　platform,　which　can　contribute　to　the　more　efficient　performance　optimization.　In　this　paper,　we　describe　the　data　processing　model　in　the　Spark　Streaming,　the　process　can　be　divided　into　multiple　phases.　We　propose　a　simple　yet　useful　method　to　explore　performance-sensitive　component　components　among　these　phases.　Experimental　results　show　that　the　proposed　method　is　suitable　for　a　wide　range　of　workloads.　At　last,　we　demonstrate　a　detail　example　of　the　application　of　this　method　on　the　typical　Spark　Streaming　workload　Word　count　and　prove　its　practicability.