Multi-drive storage planning
Putting several 32TB drives into RAID creates a very large storage pool, but raw capacity is not usable capacity and RAID is not a backup. Large arrays also make rebuild time, fault-domain size and replacement planning more important.
Quick answer
Choose RAID from the availability requirement and the recorder or server’s supported modes, not from the highest percentage of usable capacity. RAID 5 is capacity-efficient but leaves only one-disk fault tolerance; RAID 6 trades another drive for two-disk parity; RAID 10 uses more capacity for mirroring but can offer simpler rebuild behavior.
Exact models in this guide
Seagate SkyHawk AI
ST32000VE000AI-enabled NVR and edge video analytics
Interactive planner
32TB RAID usable-capacity planner
Raw capacity is only the starting point
Four 32TB drives provide 128TB of raw decimal capacity. The RAID level then consumes part of that capacity for mirrors or parity before formatting and filesystem overhead.
Always order enough bays to satisfy the protected usable-capacity target, not merely the raw total.
Raw-capacity calculations should be written down before the array is created because changing RAID level later can require destructive reconfiguration. Start with the retention target, then work backward to the protected usable capacity and number of bays. This avoids discovering after installation that parity has consumed the storage needed to meet policy.
RAID 1 with 32TB drives
A two-drive RAID 1 mirrors the data, so 64TB raw becomes roughly 32TB usable before filesystem overhead. The benefit is a complete second copy inside the array, not extra recording capacity.
Some systems support more than two mirrors in different layouts, so use the controller’s exact implementation.
RAID 1 is simple to understand but expensive in capacity. For a two-drive surveillance recorder, that trade can be acceptable when uptime matters more than density. Remember that both members contain the same logical data, so a deletion or corruption performed by the host can be mirrored instantly and is not protected by RAID.
RAID 5 capacity
RAID 5 uses the equivalent of one drive for parity, so an array of N equal drives has roughly N minus one drives of raw usable capacity.
With very large disks, rebuild time and the single-drive fault-tolerance window deserve careful consideration.
RAID 5 provides attractive capacity efficiency, which is why it appears in many storage products. With very large drives, however, the array spends substantial time reading surviving disks during rebuild. The decision should account for controller guidance, workload during rebuild, spare availability and the business consequence of another failure before reconstruction completes.
RAID 6 capacity
RAID 6 uses the equivalent of two drives for parity and can tolerate two drive failures in the array. The capacity cost is higher, but many operators prefer the extra fault tolerance for large multi-drive pools.
The controller, recorder or NAS must explicitly support the chosen drive count and RAID mode.
RAID 6 uses more raw capacity but gives the array two parity-drive equivalents. That can provide additional fault tolerance during long rebuild windows. The cost is fewer days of retained footage for the same bay count, so compare the protected capacity with the retention policy rather than judging the RAID level only by resilience.
RAID 10 capacity
RAID 10 mirrors pairs and stripes across the mirrors. With an even number of equal drives, usable raw capacity is roughly half of the total.
It can rebuild from a mirror partner without recalculating parity across every block, though actual recovery behavior depends on which disks fail.
RAID 10 trades half the raw capacity for mirrored pairs and striping. It can be attractive where predictable rebuild behavior and performance matter. Failure tolerance depends on which members fail, so “two disks can fail” is not universally true for every two-disk failure combination in a four-drive RAID 10.
Rebuilds on 32TB drives are substantial
A rebuild may need to read or write tens of terabytes while the recorder continues operating. Throughput, workload, drive health and array activity all influence how long that takes.
Do not use a simple “hours per TB” rule as a guarantee. Monitor the array and follow the controller vendor’s guidance.
Rebuild duration is not a fixed number because the controller may throttle reconstruction to preserve recording performance, and every surviving disk must share workload with live cameras. Treat a simple size divided by transfer rate as a best-case lower bound, not an operational promise. Monitor rebuild progress and temperatures in the real array.
A hot spare is not extra parity
A hot spare can reduce the time before a rebuild begins after a failure, but it does not increase the number of simultaneous failures the RAID level can tolerate before reconstruction starts.
Budget the spare as an additional drive and bay.
A hot spare shortens the delay between a detected failure and rebuild start when the controller activates it automatically. It still consumes a bay and purchase budget while contributing no normal usable capacity. Include that trade in the cost model instead of presenting the spare as “free insurance.”
RAID does not replace backup
RAID protects against some disk failures. It does not protect against recorder theft, deletion, ransomware, fire, controller corruption or a mistake that affects the whole array.
Critical footage may need replication or export to a separate system.
Backup strategy should be based on footage importance. If losing a week of ordinary recordings is acceptable, RAID may be sufficient for availability. If certain evidence must survive theft, fire or operator error, export or replicate it outside the array. The backup policy can be selective rather than duplicating every retained video block.
Retention changes with RAID
Every terabyte devoted to mirrors or parity is a terabyte that cannot hold additional video. A protected array therefore retains fewer days than the same number of drives used as independent disks.
Use the recording-time calculator after determining RAID usable capacity.
Retention math should use post-RAID usable capacity and a sensible filesystem reserve. Once that value is known, feed it into the bitrate model. This two-stage approach separates array protection from video consumption and makes it easier to explain why eight 32TB drives do not provide eight times the retention of one drive in every RAID layout.
Match drives and firmware
Using the same qualified model and capacity across an array simplifies performance expectations, replacement planning and support. Mixing unrelated 32TB models can introduce different firmware and workload behavior.
For a SkyHawk AI array, verify ST32000VE000 support in the exact storage platform.
Matching drive models simplifies spares and performance expectations, but replacement availability can change over the life of an array. Keep the minimum acceptable replacement criteria documented. If a later exact model is substituted, verify controller qualification, sector behavior and capacity rather than assuming every 32TB drive can replace ST32000VE000 without review.
Plan replacement inventory
High-capacity model availability and pricing change. Operators with critical arrays may keep a compatible spare on site rather than waiting for a replacement after a failure.
Record the exact model, firmware and warranty details when the system is commissioned.
On-site spares are valuable where replacement shipping is slow or footage availability is critical. The spare should be stored in appropriate environmental conditions and periodically checked against current firmware or support guidance. A spare that cannot be recognized by the controller when finally needed provides little operational value.
Use the calculator as a planning model
Enter drive count and RAID level to estimate raw usable TB and TiB. Then subtract filesystem reserve and compare the result with retention requirements.
The controller or recorder may use vendor-specific overhead, so treat the result as planning guidance rather than an exact formatted-capacity promise.
Use the calculator to estimate raw usable capacity, then compare the result with the exact recorder’s reported pool after creation. Vendor metadata, filesystem structures and reserve policies can reduce the final number. Capture that final formatted capacity in commissioning documents so future retention calculations start from the system’s real value.
Current exact-model offers
Live buying options for the models above
Only listings containing the exact model number can appear. Generic capacity-only matches, accessories, systems and mixed-model listings are excluded.
Current manufacturer references
Exact models, warranty terms and product specifications can change. These links are included so you can verify the current manufacturer record before purchase or deployment.
Questions people ask
Frequently asked questions
How much usable space do two 32TB drives provide in RAID 1?
Roughly 32TB raw usable before formatting because one drive mirrors the other.
How much usable space do four 32TB drives provide in RAID 5?
Roughly 96TB raw usable before filesystem overhead.
What about four drives in RAID 6?
Roughly 64TB raw usable before formatting.
What about four drives in RAID 10?
Roughly 64TB raw usable before formatting.
Is RAID 5 safe for 32TB drives?
It offers one-disk fault tolerance, but rebuild risk and platform guidance should be evaluated carefully for very large arrays.
Does RAID 6 survive two drive failures?
RAID 6 is designed for two-disk parity, subject to the controller and the failure pattern.
Is RAID 10 always safer than RAID 6?
No. They have different capacity, performance and failure characteristics.
Does a hot spare increase usable capacity?
No. A spare waits to replace a failed drive and does not store normal array data.
Does RAID count as a backup?
No. Maintain another copy for footage that cannot be lost.
Should all drives be the same model?
Using the same qualified model and capacity generally simplifies array behavior and support.