Intel® Virtual RAID on CPU (Intel® VROC) RAID Write Hole (RWH) Closure

The RAID Write Hole (RWH) is a fault scenario, related to parity-based RAID. This fault scenario can be encountered when a system power failure/crash occurs at the same time as (or close to) a drive failure. When a system crashes and drive failures are correlated events, they can lead to silent data corruption or irrecoverable data loss. Due to the lack of atomicity of write operations across member drives in parity-based RAID, a power failure during the parity write process of an active stripe may no longer correspond with the rest of the strip data. Data on such inconsistent stripes does not have the desired protection, and this can lead to invalid RAID volume corrections (silent data errors).

Intel® VROC supports the ability to close the RWH scenario in RAID 5 configurations using a feature called RWH Closure. This applies to Intel® VROC enabled platforms.

The Intel® VROC RWH Closure feature eliminates such vulnerability by using Partial Parity Logging (PPL) to non-volatile memory of the member disks of the RAID 5 volume. With hardware RAID and battery backed DRAM; logging could be used to recover. With Intel® VROC, a journaling drive can be added to preserve the partial parity and reduce the potential data loss issue.

It is important to realize that when this feature is enabled:

• The system may experience a performance impact due to the extra writes.
• The volatile cache of devices, which do not have power loss protection, Power Loss Imminent (PLI) feature, will be turned off to prevent possible data loss that may be caused by a surprise system power loss or a drive hot removal. If a device's volatile cache cannot be turned off, RWH Closure will not be enabled.
• The RAID volume rebuild times may increase because of the device cache being disabled.

The Intel® VROC driver will refrain from rebuilding a RAID 5 volume before or during the RWH recovery process.

If the RWH condition occurs during a recovery process and the recovery has not completed, Intel® VROC will attempt to resume the recover process from where it was interrupted.

If the RWH condition occurs during the process of the operating system going into hibernation mode (S4 power state), the RHW recovery process will be able to fix the failure condition for all the data being written during hibernation.

RWH Closure Protection Modes

There are two available modes of Intel® VROC RWH protection:

• Distributed: The RWH journal is stored on RAID member drives and there is no need for any additional drive. This mode provides a full protection against the RWH scenario, but introduces a performance penalty for write-intensive workloads.
• Journaling Drive: The RWH journal is stored on a separated journaling drive. That drive cannot be used for any other purpose. The performance penalty for write-intensive workloads depends on the performance of the journaling drive, but typically the penalty is lower compared to the Distributed mode.

The journaling drive needs to be at least as big as the smallest drive in the RAID volume (it will dictate the maximum size of the RAID volume). For RWH protection, Intel recommends working in Distributed mode (where the journaling information is stored across all three RAID member drives) and not Journaling Drive mode (where journaling information is stored on a single drive). Journaling is different than parity. Journaling happens first, even before the data is committed to the RAID 5 volume, and once the RAID 5 volume is ready, the journal is no longer needed. The reason the journaling drive should be at least as big as the smallest drive member in the RAID 5 volume is due to endurance considerations. While journaling does not actually require so much space, it does need to be written over again and again, requiring good endurance/capacity, and is what could make distributed journaling more effective. Distributed journaling utilizes the Power Loss Imminent (PLI) functioning of the drives, which is in turn using memory on the drive to store data instead of NAND, making endurance not a factor or concern anymore for journaling. The other option is to use a dedicated journaling device with much lesser capacity drive but with much higher endurance (as a dedicated journaling solution).

RWH Closure in Pre-OS

The Intel® VROC UEFI drivers support the ability to recover from the RAID 5 volume invalid state caused by an RWH condition for all enumerated RAID 5 volumes during system boot.

To enable RWH Closure with Journalig Drive options, the user needs to complete the create process with RWH Closure disabled, then go back one page. At this point, the user selects an available disk and then sets its properties to Journaling Drive.

RWH Backwards Compatibility

For RAID 5 volumes on drives created in previous versions of Intel® VROC, Intel® VROC will be able to leverage the older RWH Closure mechanisms. Previous Intel® VROC mechanisms implemented to address the RWH condition encompassed a combination of Dirty Stripe Journaling and Partial Parity Logging. This implementation only partially closed the RAID Write Hole. With newer Intel® VROC releases, the RWH solution included will completely close this condition (when RWH Closure is enabled). When RWH Closure is disabled, the old implementation (using Dirty Stripe Journaling and Partial Parity Logging) is used.