Empirical Measurements of Disk Failure Rates and Error Rates

摘要: The SATA advertised bit error rate of one in 10 terabytes is frightening. We moved 2 PB through low-cost hardware and saw five disk read events, several controller failures, many system reboots caused by security patches. conclude that uncorrectable errors are not yet a dominant system-fault source - they happen, but rare compared to other problems. also UER (uncorrectable rate) the relevant metric for our needs. When an happens, there typically damaged storage blocks (and errors.) Also, some may be masked operating system. more meaningful data architects Mean Time To Data Loss (MTTDL.)