"There are many examples of systems that tried and failed to implement fault-tolerant or distributed computations using ad hoc techniques rather than a transaction concept. Subsequently, some of these systems were successfully implemented using transaction techniques. After the fact, the implementers confessed that they simply hit a complexity barrier and could not debug the ad hoc system without a simple unifying concept to deal with the exceptions. Perhaps even more surprising, the subsequent transaction-oriented systems had better performance than the ad hoc incorrect systems, because transaction logging tends to convert many small messages and random disk inputs and outputs (I/O) into a few larger messages and sequential disk I/Os."
