There's at least 3 copies of all the data and possibly more spread around the world at different Tier1 and Tier2 centers. The T1 and T2 centers are very likely using different filesystems and everything is checksummed pretty much at every major step (data generation, before and after any network transfers, before analysis, etc).