One of the most meaningful recent innovations in data protection is building protection systems with Hard Disk Drives (HDD) as the storage medium. The economies and efficiencies offered by HDD are unparalleled to those of tape drives. Today, petabyte-scale appliances are built using HDD, relying heavily on RAID/Erasure Coding technologies to provide data availability.
Identifying and overcoming data protection roadblocks
This journey has not been without roadblocks. Several years ago, one of the largest suppliers of hard drives advised:
“Drive XXXX may be inaccessible after a power cycle.”
This was due to a firmware bug that would have rendered a disk drive useless – an event that could have invited significant chaos, effecting major enterprises using appliances with the impacted hard disk drives. These organizations had no idea when their storage might be unavailable and were forced to copy data back to tape drives or move to appliances using a different drive than the impacted one. There was immense amount of data lost or made unavailable, and enterprises suffered notable economic losses.
Developing data protection solutions, guided by future enterprise needs
Fast forward to today, and the situation has worsened as enterprises are building bigger systems using HDD. All HDD-based data protection systems use some sort of a RAID or Erasure Coding technology to allow for data availability when a disk fails, by providing 1 or 2 spare drives per drive enclosure. These technologies ensure all drives wear equally. Now, when one drive fails, what are the chances of other drives failing? The chances are high; one or two spare drives don’t support a system composed of 8 or more usable drives. In most cases, the remaining drives fail even before the RAID reconstruction is complete on the first failed drive.
So, how do you detect or even prevent such catastrophic failure of your protection storage? The inbuilt HDD logging and SMART tools prove to be useless in predicting such failures. While one could infer something with the number of media errors from a hard drive, that is only one kind of error and will eventually lead to false positives, increasing the support costs and system downtime.
Veritas Predictive Insights was developed to accurately predict this and other types of failures that a data protection system can face, leading to unplanned downtime. The driving concept is simple – over 15k+ NetBackup appliances are sending AutoSupport information to Veritas. We create a data lake of this information and write proprietary machine language algorithms tuned for data protection. These algorithms are always active – learning based upon the incoming information.
The success of any machine learning algorithm depends on the quality of data employed
With 15K+ appliances delivering data from the world’s leading enterprises, the quality of Veritas Predictive Insights data is clear; we are able to accurately predict data protection issues using the power of this technology.
When next firmware bug hits the HDD vendors, the #1 data protection company in the world – Veritas Technologies – will be able to preemptively identify and address the event with Predictive Insights, preventing downtime. To learn more about Veritas Predictive Insights, please visit veritas.com/pi
