Every post house that grows past two editors eventually gets the same advice: a NAS is fine for starting out, but serious facilities run a SAN. It usually arrives from someone who racked Fibre Channel in 2009, and it comes with a five-figure invoice attached. Before you take it, it’s worth pinning down what those two acronyms mean in 2026, because the market that produced that advice no longer exists. The company that defined shared storage for post now describes its flagship as a NAS, on its own spec page.
Nobody wants a SAN. What you want is four editors in one project without anyone stepping on anyone else’s bins. That is a locking problem, not a bandwidth problem, and which storage architecture you need is downstream of how your NLE handles locking. This post is the concepts and the decision. When the answer turns out to be “buy a NAS,” which it usually does, the builds live in our editing server guide.
TLDR
If you just want the answer:
- SAN means block-level storage over Fibre Channel or iSCSI. NAS means file-level storage over SMB or NFS. Neither term names the thing you’re shopping for, which is safe simultaneous writes.
- Post facilities bought SANs for guaranteed per-client bandwidth and for Avid’s collaboration semantics, not because block storage is faster. Avid dropped Fibre Channel two product generations ago, and its current spec page files NEXIS under “High-performance NAS for video editing.” Every current NEXIS engine connects to its clients over Ethernet.
- One 10GbE link carries about 1.1 GB/s in the real world. A four-editor mixed-codec house needs 125 to 375 MB/s. The throughput argument for a small-shop SAN is over.
- If you cut in Premiere or Resolve, buy a NAS and stop thinking about it. Both do their own locking. If you cut in Avid, buy the NAS and add bin-locking software at $299 per activation instead of buying a fabric.
- Real SAN territory starts where a staff engineer does: eight or more clients on uncompressed or 8K, per-client bandwidth written into a contract, and Mac, Windows, and Linux sharing one namespace.
Block vs file, in editor terms
A SAN hands your workstation raw disk blocks and trusts your machine to manage them. Your Mac formats the volume, runs the filesystem, and believes it’s talking to a local drive. The transport is Fibre Channel or iSCSI, which is the same block traffic carried over ordinary TCP/IP. A NAS keeps the filesystem to itself. Your machine asks for a file over SMB or NFS, and the box on the other end handles allocation, permissions, and locking before a single byte comes back.
That division of labor sounds like plumbing trivia. It’s the entire subject. Whoever owns the filesystem owns the answer to the question that matters in a shared edit bay: what happens when two people write at once?
The one-writer rule
The acronym explainers all skip this part. A block volume has exactly one legitimate writer unless something coordinates the writes. Mount the same LUN from two hosts with an ordinary filesystem on it, NTFS or APFS or anything else, and you will corrupt it. Not might. Neither host knows the other is allocating blocks, so they overwrite each other’s bookkeeping and the volume dies. Microsoft’s own guidance on iSCSI says it plainly: two uncoordinated hosts on one LUN is close to a guarantee of filesystem corruption.
Which brings up the setting in your NAS you may have already found. Every current QNAP and Synology ships an iSCSI target, and a persistent bit of forum wisdom says switching it on turns your NAS into a SAN. It does not. It gives one computer a network-attached virtual disk that no other computer can safely touch. Fine for a Time Machine target or a VM datastore with one host. Useless for four editors in one project, which was the whole point.
What makes a SAN shareable is a piece most people have never heard named: the metadata controller. In a real post SAN, workstations get block access to the same volume while a separate server arbitrates who owns which regions of it and holds the file locks. That’s what StorNext is, and what Apple’s Xsan was: a cluster filesystem sitting on top of the block storage, doing the coordination the storage itself can’t. A “post-production SAN” was never just the hardware. It’s SAN hardware plus a cluster filesystem plus the license fees and the care both require.
Why post ran on SANs for twenty years
Three reasons, in the order they mattered. First, bandwidth you could count on. When the office LAN was gigabit, a 2Gb or 4Gb Fibre Channel link was the only way to feed uncompressed SD or HD to a workstation, and FC is a lossless fabric on its own dedicated network, so storage traffic never fought the render farm or somebody’s email sync. Second, Avid. Media Composer’s collaboration model, project sharing plus bin locking, was built against Avid’s own shared filesystem, and an ordinary volume did not provide it. Unity, then ISIS, then NEXIS were the reference behavior, and facilities bought the storage to get the workflow. Third, one namespace across platforms: Mac finishing suites, Windows and Linux VFX, and Linux transcode nodes all mounting the same volume at the same paths, at speed. Xsan and StorNext existed for exactly this.
All three reasons have aged, and not gracefully. Xsan’s management GUI went away with macOS Server, and integrators now openly market “Xsan transition” services to move facilities off it. And Avid, the company whose workflow justified more Fibre Channel purchases than any other, left the fabric behind on its own: Unity used fibre, ISIS moved to copper Ethernet, and NEXIS has been Ethernet-only from the start.
What changed
Ethernet got fast and cheap, is what changed. A single 10GbE link moves about 1.1 GB/s in practice, and the hardware to run it costs less than a set of decent timeline drives; we covered the switches and adapters in the 10GbE guide. Against that, run the stream math from the server guide: ProRes 422 HQ at 4K60 is about 177 MB/s per stream, and a three-to-four-editor house cutting mixed codecs draws 125 to 375 MB/s in aggregate. One 10GbE port covers that with headroom. 25GbE, at roughly 3.1 GB/s, covers the shop that would have needed 8Gb Fibre Channel. There is no bandwidth case left for a small facility to buy a fabric.
The protocol grew up too. SMB3 Multichannel opens several TCP connections per session, so a client can pull aggregate bandwidth across links and survive one of them dying, without you touching bonding or LACP; macOS has shipped it enabled since 11.3. One caveat from experience, because this is where Mac-heavy shops get burned: the macOS SMB client is the weak link in the chain. There are reproducible cases of a Mac Pro on 100GbE stuck around 3 GB/s because the client opened a single TCP socket, and write-side bottlenecks against servers with Multichannel confirmed active. When a Mac room feels slower than the spec sheet promises, the ceiling is almost always the SMB client, not the disks and not the switch.
And then there’s the exhibit that closes the argument. Avid’s current spec page for NEXIS carries the title “High-performance NAS for video editing,” and heads its product list “NAS for video editing and audio production.” Those are Avid’s words about Avid’s product. The connections down the whole line are Ethernet: 40 and 50GbE on the F5, 10/25GbE on the F2 and on the PRO+. The string “Fibre Channel” appears nowhere on the page. The most SAN-coded brand in post-production ships a NAS and says so. If someone tells you a NAS can’t run a real facility, Avid disagrees, in writing.
What your NLE needs from storage
This is the real decision axis, and the one the generic SAN-vs-NAS pages never touch. Multi-editor workflow lives or dies on locking, and each NLE solved locking in its own layer.
Premiere Pro does it with project files. A Production is a folder of small projects on shared storage; whoever opens one gets the lock, everyone else gets read-only until it’s released. The storage’s job is to be a plain SMB share, and Adobe’s guidance is SMB specifically for central storage. No block semantics anywhere in the design.
DaVinci Resolve doesn’t even do it in storage. Collaborative projects live in a PostgreSQL database on a project server, and bin and timeline locks are rows in that database. The media can sit on any share all the editors can reach. Whether the bytes arrive as blocks or files could not matter less to Resolve; the locking happens in Postgres.
Media Composer is the exception that built an industry. Its bin locking expects the behavior of Avid’s own filesystem, and a plain SMB volume doesn’t provide it out of the box. This is the one workflow that ever needed the storage layer’s help, and it’s the source of every “you’ll outgrow your NAS” warning you’ve ever heard. Even here, though, the fix has shrunk from a rack to a download.
Final Cut Pro mostly opts out: libraries on network volumes have a long unsupported-or-fragile history, so FCP shops work local or DAS and share through a MAM if they share at all.
Add it up and the question “do I need a SAN?” collapses into “do I cut in Avid, and with how many people?” Two of the three major collaborative NLEs bring their own locking and just want a file share. Only one ever cared what the storage was, and its own manufacturer now sells Ethernet file storage to serve it.
The middle path: locking as software
If the missing piece is Avid’s locking behavior, you can buy that behavior by itself. Hedge Mimiq is bin locking for Media Composer on storage Avid never blessed: SMB, AFP, NFS, a NAS you already own. It’s $299 per activation as a perpetual license with no subscription, and if you only need it for one job, you can rent it at $59 for a month, the way you’d rent a camera. For a three-seat Avid room, that’s the difference between a software line item and a storage migration. The open-source route is MediaHarmony, a Samba module that gives each user their own copy of Avid’s metadata files; it’s free, and it’s also a Samba build you maintain yourself, so price the sysadmin hours in.
One rung up sit the media-storage vendors who decided the SAN-or-NAS question was worth dissolving instead of answering. SNS pitches its EVO line with the sentence “Unlike generic SAN or NAS systems,” and emulates the Avid filesystem so Media Composer sees native project and bin sharing on an Ethernet box. Facilis still answers the transport question with “yes”: the same HUB chassis offers dual-port 16, 32, or 64Gb Fibre Channel and dual-port 10 through 100GbE, with the company’s own shared filesystem doing the arbitration either way. Tiger Store lets some clients mount a volume over the fabric and others over the LAN at the same time. EditShare’s EFS runs its own client protocol for contention control across a cluster. Different vendors, one message: the locking layer is the product, and the wire is a checkbox.
The remote-first version of the same idea is LucidLink, which is what distributed teams bought instead of a SAN. It presents cloud object storage as a shared filespace with local-feeling performance, priced per member per month. Below roughly 25 TB of active project data it’s a credible shared-storage tier on its own; above that, LucidLink’s own guidance points you at their sales team, which tells you where its lane ends.
The cost reality
Look at what Avid publishes and a pattern jumps out. The NEXIS PRO+, the entry engine, delivers up to 700 MB/s through a 10/25GbE port. The F2 delivers up to 480 MB/s through the same class of port. Both figures sit comfortably under the 1.1 GB/s a single 10GbE link carries, which means Avid fits a 25GbE-capable interface to engines that can’t saturate 10GbE. Street pricing for a PRO+ 40TB runs $16,643, and F2 engines list at $23,995 for 60TB, $25,995 for 100TB, and $27,295 for 140TB, hardware only. At those prices you are not buying throughput. You are buying Avid’s collaboration semantics, certified compatibility, and a support contract with a phone number on it. For a broadcaster with a playout deadline, that contract is worth every dollar. Just be clear about which item the invoice is for.
Elsewhere in shared storage, pricing itself is the tier signal. SNS, Facilis, EditShare, and Tiger publish no prices at all; every path ends at a quote form. When a category is quote-only, the vendors are telling you who they sell to, and it isn’t a three-editor shop comparing boxes on a Tuesday. A true Fibre Channel build adds its own stack before a frame moves: an HBA in every workstation, an FC switch, optics per port, a metadata controller, and per-seat cluster filesystem licenses. Nobody will quote you that total in public either.
Against all of that stands a $299 download that gives a NAS the one behavior small Avid shops were told required the invoice above. That gap is the comparison that decides most of these purchases. Block versus file barely enters into it.
Where a SAN still makes sense
None of this means SANs are dead. It means the line moved, and it now sits where a staff engineer does. You’re in real SAN territory when most of this list is true:
- Eight or more concurrent clients working in uncompressed formats or 8K, where per-seat draw is measured in gigabytes per second, not hundreds of megabytes.
- Deterministic per-client bandwidth is a contractual requirement, not a preference. Dailies for a studio show can’t buffer because a render job woke up.
- Mac, Windows, and Linux machines need one namespace with identical paths, at full speed, simultaneously.
- Someone is employed to run the storage. A cluster filesystem with a metadata controller is infrastructure with a pager attached, not an appliance.
If that describes your facility, you already have an integrator on retainer and this article is a Tuesday refresher. If it doesn’t, the decision tree is short. Premiere or Resolve: buy the NAS. Avid, under eight seats: buy the NAS and the $299 license. We priced the actual tiers, with specific boxes at each headcount, in the small post house shared storage guide.
FAQ
My NAS has iSCSI. Do I have a SAN?
Technically yes, functionally no. iSCSI is block storage, so the definition is satisfied, but a LUN formatted with a normal filesystem supports one writer. Without a cluster filesystem coordinating access, a second host mounting that volume read-write will corrupt it. For shared editing, keep the NAS on SMB.
Can I run Avid bin locking on a Synology or QNAP?
Yes. Mimiq handles the bin locking over SMB, and the storage side needs its oplocks disabled on the share, which every serious NAS allows. OWC even publishes a step-by-step for Avid bin locking on its Jellyfish line. When NAS vendors write the how-to themselves, the gap is closed.
Is Fibre Channel faster than 10GbE?
For a single client’s block latency, marginally, because FC is lossless and skips the file-protocol overhead. For your timeline, no. Playback cares about sustained throughput, and 10GbE delivers more of it than most shared arrays will hand any one client anyway. You will not see the difference in a scrub.
Is 25GbE the new 10GbE?
It’s heading that way at the facility level, and NEXIS engines already ship 25GbE-capable ports. For a small shop, 10GbE remains the sane default: the math above shows a four-editor room fitting inside one link, and 25GbE switching still costs a multiple of 10GbE for headroom you’d never touch.
What about Thunderbolt storage?
Thunderbolt is DAS: very fast, and for one seat. The Thunderbolt “NAS” hybrids run IP over the Thunderbolt cable, so your Mac sees a network share with a fat pipe, not a block volume. Useful boxes, but they’re NAS semantics in the terms of this article, and they don’t change the decision.
In conclusion
SAN vs NAS was a hardware question in 2009. In 2026 it’s a locking question wearing an old costume. Your NLE either brings its own locking, and wants nothing fancier than a well-fed SMB share, or it’s Media Composer, and the missing behavior costs $299. The fabric, the metadata controller, and the engineer come back into the picture at a scale most of us will visit on a facility tour and never own.
And that’s it. If your next step is picking the box, the server guide covers the builds and the shared storage guide covers the tiers by headcount. If you’re staring down a quote with a fabric in it and a team of four around you, ask the integrator one question first: which of my applications needs block storage? Sit back and enjoy the pause. Questions about your setup? Leave a comment.
