Backblaze operating more than 80 thousand HDDs announced the 2017 Q 1 report of HDD failure rate, and what is the failure rate for expensive enterprises introduced massively?

Cloud storage serviceBackblazeThe first quarter report of the 2017 1st quarter report of the hard disk (HDD) failure rate which is the usual was released. Since this time, we have introduced a large amount of Seagate's enterprise version of HDD, and attention is also drawn about how much the failure rate differs from general-purpose products.

2017 Hard Drive Failure Rates - What the Numbers Tell Us

Backblaze stores and publishes the failure rate for a large number of HDDs used for cloud storage service operation for 4 years. Although it is a special use as a cloud storage servicer, in the open material of Backblaze, it is attracting attention as a reference material for general users who use HDDs because it is possible to pursue the trend over time for the failure rate of a large number of HDDs by product number I will.

In the year 2017 Q1 (January to March), Backblaze added 15,57 HDDs, and when we replace the HDDs replaced or replaced, a large amount of HDDs totaling 82,516 units were operated. All HDDs are 3.5 inch models.

In 2017, the majority added in Q1 is a 4 TB model, followed by a large capacity 8 TB model. At the moment we introduce a large quantity of 4TB products with the best unit cost of capacity at the moment and we are shifting to a large capacity 8TB model, and replacement from the 3TB model which was the mainstream in the past has been attempted.

And data on HDD failure rate by manufacturer / product number during Q1 of 2017 are as follows. The number of failures of "Drive Failures" in the table, and the failure rate when "Annualized Failur Rate (AFR)" converts to full year. Seagate's "ST 4000 DX000" shows a high AFR of 35.88%, but this also means that there are fewer than 170 units introduced, as well as "bathtub effects" that errors are likely to occur immediately after introduction as well as after aged deterioration , The year-to-year failure rate on a product life span will be 7.5%. Since it is only 170 units and the introduction number is small, Backblaze says that value as statistical data at the present time is small.

The failure rate for the HDD currently in operation in all periods after April 2013 is as follows. The failure rate of HGST made HDD such as "HDS 5C 3030 ALA 630", "HDS 5 C 4040 ALE 630", "HDS 5 C 4040 ALE 640", etc. where AFR is less than 1% is astonishing level. On the other hand, HDDs with a large number of operating units can also check models with high failure rates like Western Digital's "WD30EFRX" "WD60EFRX".

And, it is very interesting that we are introducing 2459 enterprise models called Seagate "ST 8000 NM 055". Backblaze had introduced 45 ST8000 NM 055 so far, but since it was able to receive a large discount by purchasing a large quantity through negotiations with Seagate, we have brought the 8 TB products for enterprises 50 times faster than before It seems that it was decided to introduce a large amount.

By comparison with the same Seagate 8 TB model, the comparison between "ST 8000 DM 002" for consumers and "ST 8000 NM 055" for enterprises is as follows. At the moment, ST8000NM055 for enterprises has higher AFR, but many of ST8000NM055 drives are in the early stages when data writing has not yet been completed, and considering the above bathtub effect, evaluate the failure rate He said it was too early. It seems to be necessary to operate for at least 6 months, so the real value of the HDD for enterprise will be questioned will be after the next 2017 Q2 report.

According to Backblaze, ST8000DM002 for consumers is 7.2 W at idle, the average power consumption is 9 W, ST 8000 NM 055 for enterprise is 9 W at idle, average power consumption is 10 W. However, for enterprise HDDsPowerChoiceIt is possible to optimize power consumption by technology, and it also draws attention to how much effect it is in the operation state of Backblaze which uses a large amount of HDD.

in Hardware, Posted by darkhorse_log