DieHard: reliable scheduling to survive correlated failures in cloud data centers

   Abstract