Read repair chance looks to be an awesome feature. We have a pretty high cache hit rate so I would assume read repair chance would reduce a lot of bandwidth and disk activity across our cluster. Does anyone have some statistics or experiences they want to share?
Edward
