Despite all the quarantine twists and turns, data centers continue to work. And with a greater load than before, because the volume of Internet traffic has grown significantly. Which hard drives will do the job best, Backblaze once again found out. A well-known cloud storage service has released a report on HDD reliability - for the first quarter of 2020.
The number of HDDs in Backblaze is constantly increasing. At the time of the report, the company has 132 discs at its disposal. Of these, 339 are bootable, 2 are disks for data storage. The report displays statistics on disk failures of different companies and different capacities.
So, the first quarter of 2020
By the end of the reporting period, the number of accounting hard drives was 129. These are HDDs that store customer data. Data on disks that are being tested, as well as models, the number of which does not exceed 959 copies, has been removed from the statistics. After such filtering, 60 remain. The number of days worked out by the disks is 129 million. The number of failures is 764.
Company comments
Annualized failure rate (AFR) was 1,07%. This is the minimum indicator for the entire period of monitoring the operation of hard drives, that is, since 2013. For comparison, in the first quarter of 2019, the AFR was 1,56%.
During the reporting quarter, 4 HDD models from three manufacturers never failed. Drives demonstrated zero failure rate Toshiba 4 TB and Seagate with a capacity of 16 TB. But these models had a small operating time - only about 10 days in total for the entire quarter. Accordingly, in the event of a single 000TB Seagate drive failure, the AFR would have been 16% for the quarter. Under similar conditions, the AFR of 7,25TB Toshiba drives would be 4%.
On the contrary, the models HGST many more days worked, so the AFR is not as volatile here. Had the 8TB model failed, the AFR would have been only 0,4%; if the 12TB model had failed, the AFR would have risen to 0,26% QoQ. The zero bounce rate is impressive in the first, which in the second cases.
Methodology for calculating the annual failure rate
Regardless of the observation period (month, quarter, etc.), Backblaze recalculates all failures for a year. According to the following formula:
AFR = (Disk Failures/(Runtime/366) * 100
Where:
- Drive failures β the number of HDDs that failed during the observation period.
- Operating time β the number of days during which the monitoring of disk activity lasted.
- 366 - the total number of days in a year (in non-leap years, the indicator drops to 365).
Expectations and reality compared to Q2019 XNUMX
A year ago, company representatives made several predictions about what could happen by the end of 2019 (of course, in relation to HDD performance). It's time to evaluate their justice.
Forecast: Backblaze will continue to reduce the number of 4 TB drives in use, so that there will be less than 15 by the end of 000.
Reality: There are still more than 35 disks because the company did not have enough time to replace.
Forecast: Backblaze will install a minimum of 20 20TB drives for testing purposes.
Reality: the forecast did not come true from the word at all, the company did not take a single such disk for testing.
Forecast: the total amount of file space in the Backblaze DC will exceed exabytes.
Reality: and so it happened, in March the exabyte mark was passed.
Forecast: Backblaze will install and test at least one HAMR drive from Seagate and/or 1 MAMR drive from Western Digital.
Reality: nothing happened; maybe it will happen by the end of 2020.
Complete hard drive statistics since 2013
The table below contains data on disk failures that were operated by the company from April 20, 2013 to March 31, 2020.
Complete Survey Dataset .
If you only need the tables in this article, you can with data.
Source: habr.com