## Ross Lazarus, September 2016

This is a Kaplan Meier analysis of the BackBlaze hard drive reliability data, using all available data to end second quarter of 2016 from https://www.backblaze.com/b2/hard-drive-test-data.html .

Previous posts are at http://bioinformare.blogspot.com.au/2016/05/survival-analysis-of-hard-disk-drive.html and http://bioinformare.blogspot.com.au/2016/02/survival-analysis-of-hard-disk-drive.html .

I reran my scripts and got the plots shown below. It's taking a while to read all the data as there are now a very large number of drives spinning. A total of 41740623 rows were processed in about 35 minutes on my home desktop by the python script in the github repository.

Updated curves:

and

By Manufacturer:

Add caption |

Once again for me, little change is seen in the KM curves and statistics with a lot more drives and a lot more observaton time, suggesting that this statistical approach is reliable and robust, although in general we expect that more data provides better resolution.

In terms of the KM statistical tests, additional data confirms the earlier inference that there are significant differences between the manufacturer and model risk profiles over time.

Call:

survdiff(formula = sm ~ model, data = dm, rho = 0)

N Observed Expected (O-E)^2/E (O-E)^2/V

model=HGST HMS5C4040ALE640 7168 85 505.51 349.800 406.826

model=HGST HMS5C4040BLE640 8505 29 269.99 215.103 231.736

model=Hitachi HDS5C3030ALA630 4664 117 466.48 261.826 302.989

model=Hitachi HDS5C4040ALE630 2719 71 268.60 145.365 157.458

model=Hitachi HDS722020ALA330 4774 215 472.27 140.149 161.908

model=Hitachi HDS723030ALA640 1048 55 103.54 22.753 23.459

model=ST3000DM001 4707 1705 246.40 8634.322 9272.385

model=ST31500341AS 787 216 35.74 909.141 917.789

model=ST31500541AS 2188 392 157.42 349.574 363.940

model=ST4000DM000 36089 1123 1500.66 95.042 151.313

model=ST500LM012 HN 801 26 22.42 0.573 0.577

model=ST6000DX000 1915 31 77.14 27.601 28.497

model=ST8000DM002 2754 3 3.74 0.146 0.149

model=WDC WD10EADS 550 60 46.72 3.773 3.818

model=WDC WD30EFRX 1289 136 87.38 27.053 27.637

Chisq= 11353 on 14 degrees of freedom, p= 0

Call:

survdiff(formula = s ~ manufact, data = ds, rho = 0)

N Observed Expected (O-E)^2/E (O-E)^2/V

manufact=HGST 15840 120 821.8 599.348 744.193

manufact=Hitachi 13246 462 1433.5 658.440 1046.810

manufact=HN 801 26 23.6 0.242 0.243

manufact=ST 49900 3792 2255.7 1046.249 2067.849

manufact=TOSHIBA 279 12 13.6 0.181 0.182

manufact=WDC 3920 385 248.7 74.701 78.874

Chisq= 2493 on 5 degrees of freedom, p= 0