5.2 HPE MSA Software: Virtual Storage, MSA-DP+, Automated Tiering & SSD Read Cache

HPE MSA Software: Virtual Storage and Tiering

Quick Answer: MSA Gen7 uses virtual storage: drives form disk groups, disk groups join one of two pools (A or B, each owned by a controller), and volumes are thin-provisioned from a pool in 4 MB pages. The automated tiering engine moves hot pages to the Performance tier (SSD) and cold pages to the Standard (10K SAS) or Archive (7.2K midline) tier. Mixing SSDs and HDDs requires the Advanced Data Services (ADS) Suite, except when the SSDs serve only as SSD read cache.

The virtual storage hierarchy

Because a volume's pages spread across every disk group in its pool, adding capacity also adds performance, and HPE states that data automatically reflows to balance utilization when a pool is expanded.

RAID levels and MSA-DP+

MSA Gen7 supports RAID 1, 5, 6 and 10 and MSA-DP+ for virtual disk groups. RAID 0 (NRAID) is allowed only for read cache.

MSA-DP+ is a RAID-based protection level introduced with the sixth-generation MSA. HPE describes it as using erasure coding to give better performance and availability and very fast rebuilds compared with traditional parity RAID:

• Data, parity and spare capacity are distributed across every drive in the disk group. There are no idle dedicated spare drives.
• The default target spare capacity equals the sum of the two largest drives, enough to restore full fault tolerance after losing any two drives.
• After a failure, the rebuild writes to many drives in parallel, so it finishes much faster than rebuilding onto a single spare.
• Drives in one MSA-DP+ group must be the same type. Capacities may differ, but HPE recommends that the largest drive be no more than twice the smallest.
• A replacement drive inserted into the failed drive's slot joins the group automatically, and the group can be expanded incrementally.

The three tiers

A tier is the set of disk groups of the same media class in a pool. A pool can contain one, two or all three tiers.

Automated tiering

MSA automated tiering (version 2.0 in Gen7) is sub-LUN: it moves individual 4 MB pages rather than whole volumes. The controllers track access to every page. Frequently accessed pages migrate up to the highest available tier, and pages that cool down migrate down. Movement is continuous and needs no scheduling.

The SMU I/O Workload view helps with sizing the SSD tier. It shows, over a 7-day timeline, what share of the I/O is served by what share of capacity. If 80% of I/O falls on a small slice of capacity, a modest SSD tier delivers most of the benefit.

Volume tier affinity

The QuickSpecs list Virtual Tier Affinity as the MSA quality-of-service feature. A volume's tier affinity can be:

• No affinity (default): pages move purely on access patterns.
• Performance: pages are biased toward the fastest tier, useful for a latency-sensitive database.
• Archive: pages are biased toward the archive tier, useful for backup copies or rarely used data, so they do not compete for SSD space.

Affinity is a preference, not a pin. Tiering continues, and the tier capacity still limits what fits.

SSD read cache

SSD read cache is the low-cost alternative to an SSD performance tier:

• One or two SSDs are added to a pool as a read cache disk group; RAID 0 is allowed because the cache holds only copies.
• Frequently read data is copied to SSD. Writes go to the HDD disk groups, so losing a cache SSD loses no data.
• The Gen7 array supports up to 8.34 TB (7.6 TiB) of read cache.
• A pool uses either a performance tier or read cache, not both.

Licensing: what needs the Advanced Data Services Suite

The MSA Advanced Data Services (ADS) Suite is included with every MSA 2072 and optional on MSA 2070 models. It adds performance tiering, 512 snapshots and Volume Copy, and Remote Snap. HPE's rule is that an ADS license is required when HDDs and SSDs are mixed in the same system, except when SSDs are used only as read cache.

Archive tiering between 10K and midline HDDs does not involve SSDs, so it does not trigger the license.

Thin provisioning and space reclamation

MSA virtual volumes are thin provisioned: pages are allocated only when data is written. HPE lists thin provisioning, space reclamation and thin rebuild as standard "thin technologies":

• Space reclamation returns pages to the pool when hosts send SCSI UNMAP after deleting data.
• Thin rebuild rebuilds only the allocated parts of a failed drive, shortening rebuilds on partly full disk groups.
• Because pools can be overcommitted, monitor pool thresholds and alerts so the pool never runs out of space.

Host presentation

Volumes are mapped to hosts or host groups with a LUN number and access level. Present each volume through both controllers so ALUA multipathing can fail over. The MSA also supports the VMware VAAI primitives (hardware-assisted locking, copy offload and block zeroing) and LDAP authentication of administrators against Active Directory.

Common exam traps

• Tiering versus read cache: tiering moves the only copy of data and needs ADS when SSDs and HDDs are mixed; read cache holds disposable copies and does not need ADS.
• One acceleration method per pool: performance tier or read cache, not both.
• MSA-DP+ spares: the spare capacity is distributed, not a separate spare drive.
• Two pools maximum: one per controller.