O'Reilly logo
  • Avineshwar Singh thinks this is interesting:

Despite the team’s DiRT experience, the massive 1.5 petabyte recovery took longer than the two days estimated. By the morning of March 10th, only 74% of the 436,223 audio files had been successfully transferred from 3,475 recalled backup tapes to distributed filesystem storage at a nearby compute cluster. The other 1,862 backup tapes had been omitted from the tape recall process by a vendor. In addition, the recovery process had been held up by 17 bad tapes. In anticipation of a failure due to bad tapes, a redundant encoding had been used to write the backup files. Additional truck deliveries were set off to recall the redundancy tapes, along with the other 1,862 tapes that had been omitted by the first offsite recall.

From

Cover of Site Reliability Engineering

Note

Can someone give insights how this helps?