Comment The RAID industry standard for storage is RAID-6, with recovery from a double drive failure. But it's not going to be good enough as disk capacities increase, prolonging failed disk rebuild times and so lengthening the window of unrecoverable failure if a third disk fails before the recovery from a double drive failure is complete.

This point is made by Adam Leventhal of Oracle/Sun's Fishworks in a blog. He says hard drive capacity roughly doubles every year but hard drive bandwidth is pretty constant, so it takes longer and longer to write data to fill up a drive.

Other things being equal, a 500GB drive will take twice as long to write as a 250GB drive. Suppliers are now producing 2TB drives, taking four times as long to fill with data as a 500GB drive; Leventhal implying that it will take about eight hours.

Assume 3TB drives are coming, then 4TB ones, and we're looking at 12 hours and 16 hours respectively for a rebuild of a full failed disk. Every added terabyte adds four hours to the rebuild time, half a day. That's increasing the chances a third drive will fail in the rebuild period for second and first failed drive rebuilds.

Leventhal has added triple-parity RAID to Sun's ZFS filesystem, calling it RAIDz3. He suggests calling it generically RAID-7 or RAID-8 might be silly. RAID-6 is often known as RAID-DP though, so RAID-TP would seem logical. Leventhal says it too could be superseded if disk capacities keep on growing.

That has to be logically true but, if the use of 3.5-inch disks switches over to 2.5-inch drives then that would reduce failed disk rebuild times. It would also likely increase the number of drives in an array, putting us back, roughly speaking, at square one.

Triple-parity RAID-Z will be included in the next major software release for Oracle/Sun's 7000 series sometime in the third quarter of this year; in other words, in a few weeks. It's not a first though - Avante Digital had a triple-parity EasyRAID product in 2006.

We might expect triple-parity RAID to start appearing, perhaps as an option, in mainstream enterprise EMC, HDS, HDS, IBM and NetApp arrays, and third-party RAID controllers from next year. ®

Related stories

• Oracle bigs up 7000 ZFS storage line (21 September 2010)

  https://www.theregister.com/2010/09/21/oracle_7x20/

• Data Domain doubles up dedupe speed (20 July 2009)

  https://www.theregister.com/2009/07/20/data_domain_dd880/

• Dell replaces entry-level EqualLogic boxes (17 June 2009)

  https://www.theregister.com/2009/06/17/dell_ps4000/

• ETERNUS sunshine for a spotless drive (8 June 2009)

  https://www.theregister.com/2009/06/08/fujitsu_eternus_dx/

• Rorke's drift towards RAID ASIC replacement (16 April 2009)

  https://www.theregister.com/2009/04/16/rorke_nehalem_raid/

• Why Dell's PS6000 isn't quite enterprise class (27 March 2009)

  https://www.theregister.com/2009/03/27/dell_equallogic_autotiering/

• HP tells us all about EVA (10 March 2009)

  https://www.theregister.com/2009/03/10/hp_revs_eva/

• HP to spruce EVA storage arrays (19 February 2009)

  https://www.theregister.com/2009/02/19/hp_eva6400_eva8400/

• Sun beefs up modular arrays (3 February 2009)

  https://www.theregister.com/2009/02/03/sun_modded_mid_range_arrays/

• You don't know disk about storage failures (6 March 2008)

  https://www.theregister.com/2008/03/06/disk_dominant_contributor_storage_paper/