EVA6400 VRAID6 & Disk Groups?

[patrimonio]

Greetings,

I am configuring an EVA6400 and I need to make a design decision. I don't know if VRAID6 is too new or if I am looking in the wrong places, but I'd cannot seem locate any documentation on the performance of VRAID6 versus VRAID5. I am expecting, because of the mirrored parity volume that there will be an additional write penalty in addition to the storage and performance penalty. Is this the case?

Also, if I decide to postpone my decision to use VRAID6 to a later date, do I need to configure my diskgroup as "enhanced" at the time that it is created, or can I switch between the "basic" and "enhanced" modes of a Disk Group as easy as executing a command? Is there any overhead, such as storage, associated with configuring a Disk Group in "enhanced" mode (even if I never use VRAID 6)?

Any information or guidance would be very much appreciated.

Thanks & Regards,
Michael
=====

[Sudhakar Subramaniam · Accepted solution]

VRAID6 is supported on EVA4400/6400/8400 (with version XCS 09500000 onwards). It uses dual parity to provide protection from simultaneous failure of two disk drives within a parity stripe.
- Vraid6 stripes have four data blocks and two parity blocks.
- A new disk group type has been added to facilitate Vraid6 support. The new disk group type is "Enhanced Disk group".
- The "Enhanced Disk group" can simultaneously contain virtual disks of all supported Vraid types.
- The legacy disk group (Basic disk group)type supports all Vraid types except for Vraid6.
- The disk group type is set at the time of creation and cannot be changed afterwards.
Reference: XCS09500000 release notes.
A vraid6 virtual disk can sustain two disk failure from same RSS within the diskgroup. If a third disk fails before the reconstuction of previous two failed disks in the same RSS, then disk group with RAID6 redundancy transistion to INOPERATIVE STATE, even if all VRAID1 mirror pairs in the DG have atleast one functional member.
The following event message is logged in the Controller Event Log (CEL) in case of third disk fails (before reconstruct completes for previous two failures) in same RSS within a disk group.
09c95105; A Disk Group has transitioned to an INOPERATIVE state.
Reason: State: Disk group with parity redundancy is inoperative --> Disk Group with RAID6 redundancy is inoperative;
The above event indicates that the diskgroup which was previously inoperative for vdisks with redundancy vraid0 (no redundancy) and vraid5 (parity redundancy) has now transitioned to inoperative for vraid6 vdisks also

[Daniel_279]

Write Penalty VRAID-1 2
Write Penalty VRAID-5 4
Write Penalty VRAID-6 6

[Daniel_279]

example using 15k drives in disk group


VRAID-1 dg iops 10,992 pre drive iops 115
VRAID-5 6,995 73
VRAID-6 5,130
53