Panasas decides to redefine RAID
And why not?
It's too much to demand that any vendor try and make RAID storage exciting. Our hats, however, go off to Panasas for trying.
The company this week started hyping its Tiered Parity Architecture (TPA) – technology that serves as an extension to RAID. Panasas CTO Garth Gibson, one of the inventors of RAID back at Berkeley, is once again behind this advance, which once again helps customers route around disk errors.
Part of the Panasas TPA pitch centers on the idea that customers will face more disk failures as drives grow larger.
According to some industry metrics, you're likely to encounter a disk error once for every 12TBs of data read. Even with more reliable drives, that's a problem when customers start using 1TB disks on a regular basis.
Storage systems relying on RAID can survive most of these errors, although with bigger drives you again run into more potential issues. First off, recovery times can stretch to several hours as you try to reconstruct large sets of data. Secondly, with a bunch of 1TB drives, demanding customers may well suffer from a second failure during recovery – something today's RAID systems fail to handle.
Some storage vendors such as NetApp have tried to deal with this problem by adopting RAID 6 or double parity RAID. According to Panasas, the double parity approach fails because it consumes a lot of system horsepower and requires a vast amount of additional disk to perform the backup functions.
The Tiered approach has Panasas covering horizontal (traditional RAID), vertical (parity on the drive) and network parity (a guarantee that data sent from a storage system to a server or client is exactly what it should be).
Panasas likes to claim an edge in each of these technology areas.
With Horizontal Parity, for example, the company uses multiple RAID controllers to perform recovery tasks in parallel. As a result, one customer reconstructed 800GB of data in 31 minutes.
With Vertical Parity, Panasas claims some major IP advantages over rivals. It gets around the media error issue by performing RAID on the actual drive via software. The company has proved reticent to talk about the exact nature of the technology, even though it has already filed for patents.
"We don't want to go into too much detail," said Panasas CMO Len Rosenthal. "But basically there are a bunch of codes that we implement on there to detect errors at a sector level.
"If you find a media error, then you can literally fix it just like in a RAID system. Every time you write a sector, we are writing that multiple times, calculating the parity data on it."
Like the RAID 6 approach, Vertical Parity does require extra overheard in the way of disk space. Panasas contends that its overhead - 20 per cent - equals that of RAID 6. But Panasas does not require more space as the disks grow in density, while the RAID 6 crowd does.
"In RAID 6, if you want to recover from a second disk, you have to go to triple or quadruple parity," Rosenthal said.
With Network Parity, Panasas performs a complete check on data as it moves between storage boxes and server/client systems. This can help keep the information safe from things such as firmware, server component or network transmission errors.
Overall, the company is pushing a raw performance lead over rivals with data reconstruction and then basically just a better way of dealing with larger disks that are coming to market.
Existing customers will see the new RAID tools appear at no cost with the release of ActiveScale 3.2 later this year. ®