VAST says petabyte SSDs will outgrow RAID in large systems
The storage company described a simulation at SC26 that found petabyte drives can strain RAID rebuilds.

Key takeaways
- VAST Data said enclosure local declustered RAID produced a high chance of data loss over a decade.
- Global declustering and extra parity reduced the odds to 1 in 54 million in the simulation.
- VAST's own method reduced the odds to about 1 in 62 billion.
- Metadata may be the weakest part of the system, with odds of 1 in 58 million.
In this article
VAST reports the risk
Tom's Hardware reports that on 20 November 2026, VAST Data said petabyte SSD RAID may not be enough for very large flash storage. The company spoke at SC26, the Supercomputing 2026 conference. Its principal technical strategist, Glenn Lockwood, presented a simulation of a 900 petabyte system.
The simulation examined a large parallel file system. The report said enclosure local declustered RAID can lead to very high odds of data loss over a decade. The discussion focused on petabyte SSDs. VAST said these large drives would require global declustering and extra parity to prevent loss.
What the simulation found

The simulation used real failure data and was described as a discrete event simulation. It found that the chance of losing data was one in 11 across a decade. Lockwood noted that single parity is no longer common.
Double parity still left odds of 1 in 150 over a decade and was seen as advanced for some open source file systems. The report named Lustre and BeeGFS in that context. Proprietary parallel file systems used declustered RAID within an enclosure.
With global declustering and extra parity, the simulation listed odds of 1 in 54 million. VAST's own approach reached about 1 in 62 billion. The company already publishes a method beyond the minimum bar. DDN and HPE were also mentioned in relation to declustered RAID.
Why rebuild times matter
Rebuild times grow when drives get larger. A drive with 122 terabytes of capacity would need about double the rebuild time of a 60 terabyte drive under the same speed conditions. That leaves a longer window for a second failure.
In a traditional RAID 5 set, a second failure could cause data loss. The report noted that operators may still choose large drives. Many sites have power limits, and storage power use grows with bandwidth, not capacity.
The 900 petabyte simulation involved 7,920 SSDs at 122 terabytes each. It used 180 enclosures with 44 drives each. The figures show the scale of the problem.
How VAST says to protect large flash
VAST recommends system wide erasure coding with at least three parity strips. The company said global declustering plus extra parity can lower the odds. VAST also publishes quadruple parity, which goes beyond the minimum.
The company said metadata may be the weakest part of the system. Data drives had odds of 1 in 78 trillion over 10 years, while metadata had odds of 1 in 58 million. Work on the index area began roughly 12 seconds after a failure on average, compared with about two minutes for data drives, Lockwood said.
VAST's recommended fix is meant for very large flash systems. The company said operators can use global declustering and extra parity. The goal is to avoid data loss.
Sources
Spotted a mistake? See our corrections policy. Read how News by OnDevice works in our editorial policy.



