About IBM's raid5e and RAID5EE

Problem

IBM's many servers have two special RAID architectures, raid5e and Raid5ee, what are the advantages of these two raid architectures, and are these two structures worth recommending from a professional RAID data recovery perspective?

First to paste a paragraph on raid5e and Raid5ee introduction:

RAID5E & Raid5ee

Raid5e and Raid5ee are frequently mentioned technologies that support two disk failures, and IBM's storage systems use this RAID technology extensively to achieve dual-disk fault tolerance. How exactly is it implemented, and what kind of functionality does it contain?

Figure-1 raid5e

RAID5E, which is in the RAID5 of each Extent (which is the unit used to create raid in the IBM host), adds a hot standby space, such as the HS0, HSr, HSP, etc. of the Extent tail in Figure 1. If the Extent0 fails, then the remaining extents hot backup space will be used to rebuild and redistribute the data and ensure that the remaining extents is part of the RAID5. So that even a extent failure, you can immediately have a hot standby disk to replace it, and rebuild the RAID5, resulting in fault tolerance, so as to achieve the said support two disk failures.

However, it does not tolerate the failure of any two disks at the same time, and can be viewed as a variant of the RAID5 and on-line hot-standby disks (online thermal spare drives). It distributes data from I/O operations to all disks, including hot standby disks, thereby reducing the bandwidth per disk and resulting in greater efficiency. However, this means that a hot spare disk cannot be shared by multiple arrays.

In the raid5e, there is no dedicated hot standby disk, just as there is no dedicated checksum disk in the RAID5, and hot standby blocks are distributed across all disks, so that for 10 disks, 80% of each disk is used to store data, 10% for storage checksums, and 10% for hot spares. raid5e

Figure-2 Raid5ee

In addition, Raid5ee and raid5e are similar, except that the hot standby space is distributed in various extents, as distributed as the RAID5 test data; If a extent fault is found, then the hot standby space in the remaining extent (as in Figure 2, HS0, HS1, HS2, etc.), will be immediately used to reconstruct the data and ensure that it becomes part of the original RAID5 to support two disk failures.

Compared with raid5e, it does not place the hot standby space at the end of each extent, but distributes it in the data block, and it does not allow any two disks to fail at the same time. However, Raid5ee may be more convenient and flexible to address when making hot replacements.

Here are my views on these two structures:

RAID5E and Raid5ee are actually optimized raid5+hot SPARE, which is designed to allow redundant hot spares to participate in RAID groups and achieve faster parallel IO on a single raid bus.

From the raid5e and raid5ee construction Way, in addition to providing a slightly faster reading and writing IO, other advantages are not obvious, many data said that can achieve faster reconstruction speed, in fact, in raid5e or raid5ee degraded, the controller to use a very complex algorithm (relative), Implement online change RAID level (RAID5 changed to variant by raid5e or RAID5EE). At the same time, because of the raid level changes, all the hard drives in the RAID group are read and write fully. Adding a new disk will also change the RAID5 back to the original RAID structure via a complex algorithm. So in many raid5e or raid5ee cases we have contacted, the customer is not doing anything, but the data is corrupted, of course, this may be the controller firmware failure, but it is undeniable that this complex operation, high load IO is also one of the root causes.

Therefore, in the absence of specific requirements for IO performance (RAID5E and RAID5EE performance improvements are not obvious), it is recommended to use RAID5E or raid5ee as little as possible.