Is Your RAID Setup On Borrowed Time?

5 server arrays with a ticking clock above them

Posted on Tuesday, October 1, 2019

Despite improved reliability over its predecessors, RAID 5 concealed a risky design flaw. Because the design only allowed for a single redundant drive, losing a second disk would cause the complete failure of the array.

In an ideal world, the first failed disk could be replaced before a second drive could go wrong. But if the array contains drives from the same manufacturing batch, plus the additional strain placed on remaining disks, the risk of a second failure increases.

And as drives have increased in capacity, so too has the time to rebuild the RAID 5 array. Which increases the window of risk, known as the “write hole.”

Raid 6 fixed this problem, right?

RAID 6 was designed as a solution to this problem, using parity to support multiple drive failures. At the time of its release, engineers estimated that the technology decreased risk by a magnitude of 3800x over RAID 5.

But the truth is RAID 6 simply reduced risk for a while. The array rebuilding process still places additional strain on disks, and the chances of further disk failures increase accordingly. As disk capacities have continued to grow, so too has the risk that a third drive will fail, causing complete loss of the array.

In fact, according to calculations by NetApp storage engineers, 2019 would be the year that the risk of RAID 6 array failure was the same as RAID 5. This means the chances of complete array loss are now genuinely significant, as the rebuild process may take days to complete due to the capacity of disks involved.

For any business still using RAID 6 arrays, it is time to assess the health of your hard drives– and whether the technology still has a place in your data center.

For more help and advice on legacy storage, please get in touch.

Download article as a PDF - Is Your RAID Setup On Borrowed Time?

More Articles

Server cabinets in the bottom of the frame shrinking to nothing then expanding back to regular size

Cisco Storage EoSL: What Happens Next?

Like every OEM, Cisco retires older storage product lines periodically. Here we explore some of your options when this happens.

Bitcoin coming out the top of a server cabinet

Blockchain Is About To Get A Lot Bigger

Scalability is one of the biggest issues with current blockchain implementations– but that’s about to change.