SSD vs HDD RAID arrays: performance, endurance and rebuild considerations

Solid-state drives and hard drives can both sit behind the same Dell PERC or HPE Smart Array controller, but they should not be treated the same way. The vendors set rules on mixing them, recommend different cache and strip settings for SSD volumes, and give you tools to watch SSD wear. This guide collects those rules and settings so you can plan SSD and HDD arrays side by side.

Mixing rules: one media type per array

• Dell PERC 11 and PERC 12: mixing hard drives and SSDs in a virtual disk is not supported. Mixing drive speeds (7,200, 10,000 or 15,000 RPM) and link rates is supported as long as the drive type (SAS or SATA) and technology (HDD or SSD) stay the same. NVMe cannot be mixed with SAS or SATA.
• Dell PERC 12 replacement rules: a replacement must use the same media type, and an SSD replacement must have the same or a higher endurance class: read-intensive can be replaced by RI, MU or WI; mixed-use by MU or WI; write-intensive only by WI.
• HPE Smart Array: the SSA CLI guide states that mixed drives are not allowed on the same array or logical drive. Before replacing a drive, HPE says to confirm it is the same type as the degraded one: SAS or SATA, and hard drive or SSD.

You can still use both media on one controller as separate volumes, for example an SSD RAID 10 for a database and an HDD RAID 6 for file shares. For interface rules (SAS, SATA, NVMe), see SAS vs SATA vs NVMe.

Controller settings for SSD volumes

Dell FastPath. Dell describes FastPath as a feature for high IOPS on SSDs that needs the virtual disk set to write-through and no read-ahead. Dell notes that SSD virtual disks may see no difference from controller cache and may benefit from FastPath instead. Limits: only I/O smaller than the stripe size is eligible, FastPath is disabled while background operations such as rebuild or initialization run, and RAID 50 and RAID 60 virtual disks cannot use it. Cache policies are explained in write-back vs write-through, strip sizes in stripe size basics.

HPE SSD Smart Path. On Smart Array controllers, SSD Smart Path lets I/O requests bypass the RAID firmware and go straight to the SSDs. HPE says it accelerates reads for all RAID levels and writes for RAID 0, for requests that a single drive can satisfy. It is set per array with the ssdsmartpath keyword; HPE's example:

ssacli ctrl slot=0 array a modify ssdsmartpath=disable

HPE SSD over-provisioning optimization. This optional feature initializes SSDs so the whole capacity can be used for managing writes and wear levelling. It runs when the first logical drive in an array is created, and when a drive replaces a failed one. HPE lists improved SSD write performance and endurance as benefits. It is off by default in the CLI:

ssacli ctrl slot=1 create type=ld drives=1I:1:1 ssdopo=on

Monitor SSD wear

Dell iDRAC9

iDRAC9 can alert on SSD wear. In the web interface go to Configuration > System Settings > Hardware Settings > SSD Wear Thresholds:

1. Set Remaining Rated Write Endurance (1 to 99%, default 10%).
2. For NVMe SSDs, set Available Spare Alert Threshold (1 to 99%, default 10%). Dell notes Available Spare is not available for SSDs behind a PERC or HBA.
3. When a value drops below the threshold, iDRAC logs it in the Lifecycle Controller log and can send email, SNMP, IPMI, remote syslog and OS log alerts, depending on your alert settings.

racadm set System.Storage.RemainingRatedWriteEnduranceAlertThreshold 10

Set up the alert destinations as described in iDRAC9 alerts by email, SNMP and syslog.

HPE Smart Storage Administrator

HPE's SSD wear gauge reports show the current usage level and remaining expected lifetime of the SSDs on a controller. From the SSA CLI you can show an SSD summary, or add the SmartSSD Wear Gauge report to a diagnostic file:

ssacli ctrl all show ssdinfo detail
ssacli ctrl all diag file=/tmp/ssa-diag.zip ssdrpt=on

Rebuild and failure considerations

• Drive performance sets rebuild speed. HPE lists drive performance, I/O load, drive count and rebuild priority as the factors in rebuild time. Large, slow hard drives give the longest degraded windows; see rebuild time and risks.
• Background operations reduce SSD performance on Dell. FastPath is disabled during rebuilds and initialization, so expect lower SSD performance until they finish.
• Patrol read on Dell runs by default every seven days on configured SAS and SATA hard drives, and covers hot spares too.
• Drive types not supported: Dell's PERC 12 guide states that shingled magnetic recording (SMR) drives are not supported on PERC.
• Replace worn SSDs early. iDRAC's wear alert is designed so the administrator can back up or replace the SSD before it wears out.

Frequently asked questions

Can SSDs and HDDs share one controller?

Yes, as separate virtual disks or arrays. Dell and HPE prohibit mixing them inside one virtual disk or array, not on the controller.

Should SSD volumes use the controller cache?

On Dell PERC, Dell recommends write-through and no read-ahead after the initial data load so FastPath can work. On HPE Smart Array, SSD Smart Path bypasses the firmware for eligible requests.

Checklist

• One media type per virtual disk or array; separate SSD and HDD volumes.
• Replacement SSDs of the same or higher endurance class.
• Dell SSD volumes: 64 KB strip, then write-through and no read-ahead for FastPath (not RAID 50/60).
• HPE SSD arrays: review SSD Smart Path and over-provisioning optimization when creating the array.
• SSD wear alerts set in iDRAC9, or SSD wear reports checked in HPE SSA.
• Rebuild time planned for large hard drives.

Skyline can supply, configure and maintain Dell, HPE and Huawei servers under an annual maintenance contract (AMC). Contact us.