How to Build Portable RAID Systems for On-Set Backup

The first rule: build the RAID around the day’s offload workflow, not around the biggest capacity number you can afford. On set, a portable RAID is there to do three jobs: accept camera cards quickly, survive normal handling and transport, and hand clean, verified media to post without becoming the only place the footage exists. RAID isn't your backup plan by itself, but it's one storage target inside the plan. A RAID 5 shuttle can keep working after a single drive failure, but it doesn't protect you from accidental deletion, theft, a corrupted copy, a dropped enclosure, bad power, or the wrong folder being formatted. Treat the portable RAID as your fast local destination, then pair it with at least one more physical copy and, when the production supports it, an offsite or cloud copy.

Start with throughput, not terabytes

A good on-set RAID build starts with a simple sizing question: can the system ingest cards faster than camera department fills them? That depends on the number of cameras, codec, shooting ratio, card size, reader speed, and how many simultaneous copies you need to make. A single-camera interview day shooting compressed 4K doesn't need the same setup as a multi-camera reality day, a commercial with high-speed bursts, or a drama shooting ARRIRAW, REDCODE, X-OCN, or ProRes 4444 XQ. Think in sustained write speed, not advertised interface speed. Thunderbolt 3 may offer a 40 Gb/s pipe, and many pro shuttle RAIDs advertise performance well above 1 GB/s depending on configuration, but real ingest is limited by the slowest part of the chain:

Camera card reader speed
Host port and cable
RAID level
Drive type and drive count
Verification method
Number of simultaneous destinations
Thermal throttling
Filesystem overhead
Copy software behavior

The takeaway is that a RAID needs enough sustained performance for your team to copy and verify without delaying card turnover. If editorial needs to start cutting from that same unit, your performance target goes up. If the RAID is only a set-side holding volume and shuttle copies go elsewhere, reliability and power stability matter more than peak speed.

Match the chassis to the job

Portable RAID systems generally fall into a few useful categories.

Portable RAID chassis scale from small two-bay units to larger multi-bay systems.

Small rugged two-bay units are easy to carry, often bus-powered or low-power, and simple enough for lightweight shoots. They're useful for travel jobs, documentary work, small crews, or as shuttle drives. The tradeoff is limited RAID flexibility. With two drives, you're usually choosing between RAID 0 for speed and capacity or RAID 1 for mirroring. RAID 0 is risky as a backup target because either drive failing can take down the volume. RAID 1 is safer but cuts usable capacity in half. Four-bay shuttle RAIDs are the common middle ground. They're still portable, but they have enough bays for RAID 5 or RAID 10. RAID 5 gives a good balance of capacity, speed, and single-drive fault tolerance. RAID 10 gives strong performance and redundancy, but usable capacity is only half of the raw drive total. Many professional four-bay units ship preconfigured as RAID 5 because it fits the on-set and post-production use case well. Eight-bay shuttle RAIDs step up capacity, performance, and RAID options. They're better for multi-camera productions, long-form jobs, and situations where the RAID may travel from set to editorial and stay active through picture lock. With more bays, RAID 6 becomes more attractive because it can tolerate two drive failures. That matters when rebuild times are long or the array holds many terabytes of camera originals. Mobile NAS or transportable server-style units are a different category. They can be useful when multiple workstations need shared access, when the DIT cart already has networking, or when post wants the storage to arrive as a managed workspace rather than a loose drive. The penalty is complexity, which means networking, permissions, SMB behavior, power, and shutdown procedure all become part of the system. For many sets, direct-attached Thunderbolt or USB-C RAID is simpler and less fragile. For a cart-based DIT or media manager, the best chassis is usually the one your team can mount or place securely, has replaceable drives, has clear drive status indicators, supports management software, and doesn't depend on a nest of adapters to function.

Choose RAID level for failure behavior

RAID level is really a decision about what kind of failure you can tolerate.

RAID level	Minimum drives	Usable capacity	Fault tolerance	Good fit	Main caution
RAID 0	2	100% of raw capacity	None	Scratch space or an extra temporary copy when speed matters	Any single drive failure can lose the whole volume
RAID 1	2	50% of raw capacity	1 drive	Small two-bay shuttle sets and lightweight field mirrors	Capacity is cut in half and write speed is not the main benefit
RAID 5	3	Total capacity minus 1 drive	1 drive	Four-bay portable RAIDs balancing speed, capacity, and redundancy	A degraded array is vulnerable during rebuild, especially with large HDDs
RAID 6	4	Total capacity minus 2 drives	2 drives	Larger hard-drive arrays, especially eight-bay shuttles	Lower usable capacity and potentially slower writes than RAID 5
RAID 10	4	50% of raw capacity	Usually 1 drive per mirrored pair	SSD arrays, high-performance workflows, and fast rebuild needs	Capacity cost is high compared with RAID 5 or RAID 6
JBOD	1	Varies by implementation	Usually none for spanned volumes	Separate disk presentation or specific controlled workflows	Failure behavior depends on the enclosure and layout

The common options for portable media work are:

RAID 0: Maximum speed and capacity, no redundancy. Avoid for primary on-set backup unless it's only an additional temporary copy.
RAID 1: Two-drive mirror. Simple and resilient against one drive failure, but only half the raw capacity is usable.
RAID 5: Good capacity efficiency with single-drive fault tolerance. Common default for four-bay and larger shuttle systems.
RAID 6: Dual-drive fault tolerance. Better for larger arrays, especially eight-bay units, but writes can be slower than RAID 5.
RAID 10: Fast and resilient, built from mirrored pairs. Good performance, but half the raw capacity is usable.
JBOD: Separate disks or a concatenated layout, depending on implementation. Useful only when you have a clear reason and understand the failure behavior.

For most on-set backup RAIDs with four bays, RAID 5 is the default sane choice. For eight bays holding a lot of original camera media, RAID 6 deserves serious consideration. For SSD arrays where speed matters and capacity budget is less painful, RAID 10 can be excellent. The trap is treating RAID redundancy as permission to run with fewer copies. A degraded RAID isn't a healthy backup because once a drive fails, the array is under stress, and a rebuild can take hours or days on large hard drives. During that period, performance drops and the risk profile changes. If a portable array degrades on set, stop treating it as a trusted primary target until you have another verified copy elsewhere.

Hardware RAID, software RAID, or NAS RAID

Most production-friendly shuttle units use hardware RAID or enclosure-managed RAID. The host computer sees one volume, while the enclosure handles the drive set. This is usually the right choice for on-set work because it's portable between machines and doesn't depend as heavily on a specific operating system configuration. The advantages are straightforward:

Simple mounting on macOS or Windows after formatting
Vendor utility for RAID changes, drive status, and firmware updates
Clearer behavior when the unit is moved between workstations
Fewer moving parts for the media manager to explain at wrap

Software RAID can work, but it's more dependent on the host OS and configuration. A Linux-based USB RAID made from multiple external drives can be made to work in some controlled contexts, but it isn't ideal for high-pressure on-set use. USB device resets, hub behavior, drive ordering, sleep settings, and cable strain all become real failure points. NAS RAID brings its own strengths, especially for shared access and managed permissions, but it adds network setup and administration. If you need several machines pulling from the same pool, a NAS may be worth it. If your main task is fast card offload to verified copies, a direct-attached RAID is usually easier to defend. Whatever controller style you choose, install the vendor utility on the DIT workstation before the shoot. Some shuttle systems ship ready to use, often in RAID 5, but you still need the management utility for monitoring, firmware, RAID reconfiguration, and drive replacement. Don't discover on location that the only laptop on the cart can't see the RAID health state.

Drive selection is about consistency

For hard-drive arrays, use drives that are intended for RAID or NAS workloads, not bargain desktop disks. You want consistent behavior under sustained writes, vibration tolerance, and firmware that doesn't disappear for long error recovery cycles. Capacity-matched drives from the same class simplify setup and replacement.

Matched RAID-ready drives belong together; a mismatched bargain drive stands out.

Key drive criteria include:

CMR recording, not SMR, for hard-drive RAID sets
7200 RPM where sustained throughput matters
RAID or NAS-rated firmware
Adequate workload rating for repeated ingest days
Matching capacity across bays
Available spares in the same model family or approved compatibility list
Clear labeling by bay and serial number

SMR drives are especially worth avoiding in write-heavy RAID workflows. They can perform acceptably in light desktop use, then collapse under sustained ingest or rebuild activity. On set, that turns into long copy windows and unpredictable verification time. SSDs change the equation because they handle movement better, run silently, and deliver excellent random and sustained performance. They're ideal for travel, rough locations, high-speed workflows, and compact all-flash shuttles. The tradeoffs are cost per terabyte, heat concentration in small enclosures, and write endurance. For productions generating tens of terabytes a day, hard-drive shuttles may still be the more economical local copy target, with SSDs used for working media, dailies, or editorial handoff.

Power is part of the storage system

A portable RAID is only as reliable as the power feeding it. Bus-powered devices are convenient, but spinning disks need stable power, and even 2.5-inch drives can draw enough current to cause trouble when multiple devices share a hub or laptop bus.

Stable UPS-backed power supports the workstation, RAID, and ingest gear.

For AC-powered shuttle RAIDs, plan the cart like a small post room. Use clean power distribution, strain relief, and a UPS when possible. The UPS needs to give the workstation and RAID enough time to finish or pause transfers and shut down cleanly during a generator change, battery swap, or accidental unplug. Your power plan should account for:

RAID enclosure wattage under load
Laptop or workstation power draw
Camera card readers
Powered USB or Thunderbolt hubs
Monitor, router, and network switch if used
UPS runtime under real load
Safe shutdown procedure
Cable routing where crew won't kick it

Avoid hanging a RAID, multiple readers, and shuttle drives off an unpowered hub. If a hub is necessary, use a quality powered unit and test it with the exact readers and drives you'll use on set. Cheap hubs often work during a quick mount test and fail during sustained transfers. Also disable aggressive sleep settings on the workstation and understand the enclosure’s power-saving behavior. A RAID that spins down during lunch may not wake in the way your copy software expects. On a long shoot day, boring power behavior is a feature.

Ports, cables, and filesystems matter

USB-C is a connector shape, not a performance guarantee. A USB-C cable may support USB 2.0, USB 3.x, USB4, Thunderbolt 3, or Thunderbolt 4 depending on the cable and devices. Test Thunderbolt shuttle RAIDs with known-good Thunderbolt cables, not random charging cables from the pouch. The same goes for host ports because two ports that look identical on a laptop may not share the same capabilities or bus behavior. If the cart relies on simultaneous card reader input and RAID output, test the exact port layout. Some setups perform better when the reader and RAID are on different buses. Others bottleneck through a dock. Your team should decide filesystem choice with post, not guess on set. Common choices include:

APFS for modern macOS-heavy workflows
HFS+ for older macOS compatibility
exFAT for cross-platform shuttle use, with fewer professional filesystem features
NTFS for Windows-first workflows
SMB shares for NAS-based systems

If the RAID will travel from set to editorial, format it for the machines that will actually mount it. Reformatting after day one because editorial can't read the drive isn't a storage plan.

Heat and movement are normal set conditions

Most portable RAIDs are designed for desks, carts, and transport cases, not direct sun on asphalt next to a generator. Make thermal management part of the build. Hard drives and SSDs both dislike heat, but they fail differently. Hard drives are sensitive to vibration, shock, and sustained high temperature. SSDs can throttle hard when the controller gets hot, turning a fast offload rig into a slow one halfway through the day. Aluminum enclosures help, but only if air can move around them. Use the RAID in open air, not inside a closed foam case while powered, and don't stack drives tightly because it looks tidy. Keep vents clear, and keep the unit out of direct sun and away from heaters, rain covers with no airflow, dusty ground, and generator exhaust. If the chassis has fans, listen for changes. A fan clogged with dust or fabric can turn a safe array into an oven.

A powered RAID needs open air and vent space instead of a closed foam case.

Transport is separate from operation, which means a rugged case is for moving the unit, not for running it unless the case has been designed with airflow. Let spinning disks park before moving the enclosure. Secure cables so a bumped laptop doesn't yank the RAID off the cart.

Build the copy workflow around verification

The cleanest RAID build still fails the job if your team formats cards after an unverified drag-and-drop copy. Treat every camera card as being in one of three states: full, copied, or verified. “Copied” means files exist somewhere else. “Verified” means checksum verification confirmed the copy byte for byte. A reliable on-set flow usually looks like this in practice:

Camera department delivers a card, and your team marks it as full
Offload software copies to the primary RAID and a second destination
Your team creates and verifies checksums
Your team saves reports with the day’s media
Your team releases only verified cards for formatting
Your team sends one copy off location or starts upload when bandwidth allows

This is where RAID fits into the larger 3-2-1 idea: three copies, on two different media types, with one offsite or geographically separate. On a small job, that might mean a RAID on the cart, a shuttle drive for production, and a drive leaving with a producer. On a larger job, it might mean RAID plus LTO, cloud upload, or a managed transfer to editorial. Your team should verify during the workflow. If a checksum fails, you want to know while the camera card is still available.

Organize the RAID for post, not just set

Portable RAIDs often become the bridge between production and editorial, so folder structure matters. Use a clear project workspace layout that separates camera originals, sound, reports, LUTs, stills, proxies, and editorial deliverables. Keep each shoot day and camera unit distinct, and avoid dumping everything into one giant “Footage” folder and expecting assistants to untangle it later. A simple structure might include folders for:

Camera originals by shoot day, camera, and card
Production sound by shoot day and recorder roll
Offload reports and checksum manifests
CDL, LUT, and color metadata
Proxies or dailies exports
Documentation and camera reports

This mirrors how shared storage works later in post: users need to find the media relevant to their role without wading through unrelated material. A portable RAID should arrive with enough structure that editorial can ingest, relink, or clone without asking set to decode the folder names.

Common failure modes to design out

Most on-set RAID problems are ordinary problems your team didn't plan for. Watch for these failure modes during the build:

Using RAID 0 as the only fast copy
Using cheap SMR drives in a write-heavy RAID
Feeding multiple drives and readers from unpowered hubs
Connecting Thunderbolt devices with charge-only USB-C cables
Leaving the RAID management utility off the workstation
Running the enclosure inside a closed case
Working without a UPS during unstable power conditions
Formatting cards after copy but before checksum verification
Letting one person carry all copies in the same bag
Starting without a tested plan for a degraded array

The fix is to remove ambiguity, so label bays, label cables, test ports, document RAID level, save reports, and make sure everyone agrees when a card is safe to format.

A sane build for most productions

For many real productions, a solid portable RAID setup is a four-bay or eight-bay Thunderbolt shuttle, populated with RAID-rated CMR drives or SSDs, configured as RAID 5 for four bays or RAID 6 for larger hard-drive arrays. It runs on stable AC power with UPS protection, connects with known-good cables, and is monitored through the vendor utility. It receives verified camera offloads through dedicated copy software, while a second drive, LTO target, NAS, or cloud transfer creates another independent copy. Smaller productions can scale this down to a rugged two-bay mirror plus shuttle drives. Larger productions can scale it up to multiple RAIDs, NAS storage, LTO, and cloud delivery. The logic stays the same: fast enough to keep up, redundant enough to survive a device failure, organized enough for editorial, and never treated as the only copy of the footage.

FAQ

No. RAID protects against certain drive failures, but it isn't a complete backup strategy. A RAID 5 or RAID 6 array may keep working after one or more drive failures, but it won't protect footage from accidental deletion, theft, enclosure damage, corruption, bad power, or formatting the wrong card. Treat the portable RAID as one copy in a larger workflow that includes at least one additional verified copy, ideally with one copy stored offsite or geographically separate.

For most four-bay portable RAIDs, RAID 5 is a practical default because it balances usable capacity, speed, and single-drive fault tolerance. For larger eight-bay hard-drive arrays, RAID 6 is often worth considering because it can tolerate two drive failures, which is useful when rebuilds take a long time. RAID 10 is strong for performance and redundancy, especially with SSDs, but it uses only half of the raw capacity. RAID 0 shouldn't be used as the only on-set backup target because a single drive failure can take down the whole volume.

Hard drives are still cost-effective for large-capacity camera-original storage, especially when a production generates many terabytes per day. Use RAID-rated or NAS-rated CMR drives rather than SMR desktop drives. SSDs are better for speed, travel, silence, and resistance to movement, but they cost more per terabyte and can throttle if the enclosure gets hot. Many productions use hard-drive RAIDs for high-capacity verified storage and SSDs for working media, dailies, proxies, or fast editorial handoff.

A UPS is strongly recommended for AC-powered RAIDs, especially on carts using generators, temporary power, or busy locations where cables may be bumped. The UPS doesn't need to power the entire set. It should give the workstation and RAID enough time to finish or pause transfers and shut down cleanly. Sudden power loss during a copy, verification pass, or RAID write can create corruption risks and may force time-consuming troubleshooting.

Choose the filesystem based on the systems that will actually mount the RAID in post. APFS is common for modern macOS workflows, HFS+ may be needed for older Mac compatibility, NTFS fits Windows-first environments, and exFAT can work for cross-platform shuttles but has fewer professional filesystem features. If the RAID is a NAS, access may happen through SMB shares instead. The key is to confirm the requirement with post before formatting, not after the first shoot day.

Keep the RAID organized by shoot day, camera, reports, sound, and deliverables, then share only the folders or collections each person needs. Aspect supports view, download, comment, and edit controls for granular sharing permissions.