Constellation　Model　is　a　grid　resource　management　model.　It　is　a　fundamental　but　challenging　issue　for　reputation　systems　to　evaluate　target＇s　trustworthiness　accurately　and　efficiently　in　Constellation　Model.　We　present　a　reputation　model　in　which　local　feedback　is　an　aggregate　value　based　on　a　node＇s　prior　behaviors　and　performance.　The　proposed　local　feedback　metric　is　a　dynamic　one　which　changes　according　to　the　node＇s　behaviors.　We　adopt　feedback　credibility　to　evaluate　the　trustworthiness　of　a　node＇s　local　feedback.　In　our　model,　to　promote　a　node＇s　reputation,　there　exist　two　possible　ways:　the　first　one　is　to　provide　high-quality　services　to　others.　The　second　way　is　to　provide　authentic　feedbacks　towards　others.　Experimental　results　show　that　the　proposed　model　can　improve　the　evaluation　accuracy　of　global　reputations　by　leveraging　feedback　credibility　in　iterations.