A network reference build for a 2 to 5 person post team

Here is the whole post in one sentence: a 2 to 5 person edit team needs exactly one 10 GbE link to the NAS, a multi-gig switch with two real 10G uplink ports, a Thunderbolt 10GbE adapter on each Mac, and Cat6a between them, and you can build all of it for well under the price of a single year on a managed cloud suite. The hard part is not the gear. It is knowing where the 10 GbE belongs and where it is wasted money, so the budget lands on the one link that actually feeds five people at once. This is the reference build I would hand a small post house today, with the parts named and the prices dated.

The shape of a small post network #

Think of the network like a kitchen during a dinner rush. The NAS is the walk-in fridge that everyone pulls from, and the link to that fridge is the door. If five cooks share one narrow door, they queue. The trick is to make the fridge door wide (10 GbE to the NAS) and let each cook walk up at a comfortable pace (2.5 GbE per editor), because no single cook needs the full width of the door, but the door has to be wide enough that several arriving at once do not jam it.

That is the asymmetry that defines a post network and that most home-grade gear gets wrong. A single editor scrubbing ProRes 422 HQ at UHD pulls a few hundred megabits a second sustained, well inside what 2.5 GbE delivers. But three or four editors hitting the same volume at once add up fast, and the NAS link is where they collide. So the rule for a 2 to 5 person team is blunt: put the fat pipe on the server, give the desks the cheaper multi-gig ports, and buy a switch whose job is to bridge the two without becoming the bottleneck itself.

The switch is the spine, so buy the right one #

The switch decides everything downstream, so this is where to spend attention. For a 2 to 5 person team you want a small managed multi-gig switch with a bank of 2.5 GbE ports for the desks and at least two real 10G ports for the server and a future second server or NAS. Managed matters less for raw speed and more for the boring things that keep a busy volume sane: VLANs to keep a noisy guest device off the media network, link aggregation if you outgrow a single 10G link, and jumbo frames.

The switch I keep landing on is the MikroTik CRS310-8G+2S+IN, eight 2.5 GbE ports plus two 10G SFP+ cages, running RouterOS v7 or the simpler SwOS, listed around $175 to $210 depending on the retailer (checked Jun 2026). If you would rather click a web UI than touch a CLI, QNAP's QSW-M2108-2C has the same eight-2.5G-plus-two-10G shape with the 10G ports as RJ45/SFP+ combo cages, driven by QNAP's QSS web interface (checked Jun 2026). When the team grows past five and you want 10 GbE on more than one desk, QNAP's QSW-1208-8C steps up to twelve 10GbE ports, eight of them RJ45/SFP+ combo, unmanaged, at a 240 Gbps switching capacity (checked Jun 2026).

One spec to weigh before you click buy: heat and power. A 10GBASE-T (RJ45) port draws roughly 1.5 to 4 watts depending on cable length, while an SFP+ port sips about 0.7 watts regardless of distance (checked Jun 2026). That sounds like nothing until you stack eight of them in a closet with no airflow, which is why the all-copper 10G switches tend to run hotter and louder. For a small room, an SFP+ uplink to the NAS with 2.5G copper to the desks is the quiet, cool compromise. If you want the deeper switch-by-switch breakdown, that is its own piece in the post studio switch guide, and 10 GbE on a budget covers the cheapest combinations that still work.

NICs: one per Mac, one fat one for the NAS #

Modern Macs do not ship with 10GbE, so the editor side is a Thunderbolt adapter. The two mainstream picks are the Sonnet Solo10G and the OWC Thunderbolt 10G Ethernet adapter, both single-port RJ45 10GBASE-T with NBASE-T fallback down to 2.5G and 1G, both around $199 (checked Jun 2026). They are bus-powered, fanless, and pocketable, and they work on Apple Silicon and Intel Macs with Thunderbolt 3, 4, or 5. The OWC Thunderbolt 4 model advertises real-world throughput over 900 MB/s (checked Jun 2026), which is the honest ceiling of a single 10G link after overhead.

Here is the spending decision most teams get backwards. You do not actually need a 10GbE adapter on every desk. A single editor's sustained scrub rate fits inside 2.5 GbE, and a 2.5G connection runs over the Cat5e or Cat6 you probably already have, with no Thunderbolt adapter at all on Macs that have a 2.5G port or a cheap USB adapter on those that do not. Put the $199 Thunderbolt 10G adapters where they earn it: the workstation that ingests and transcodes, the colorist who hammers the volume, and the NAS itself. The NAS gets a proper 10GbE NIC or a built-in 10G port, not a Thunderbolt dongle. That is the fat pipe, and it is non-negotiable.

Cabling and where the 10 GbE physically goes #

The cable is the part people skimp on and then chase ghosts for a week. For 10GBASE-T over copper, Cat6 carries 10 Gbit only to about 55 meters, and only in a clean, low-crosstalk run; Cat6a is rated for the full 100 meter channel at 500 MHz (checked Jun 2026). In a small studio your runs are short, so Cat6 will often work, but Cat6a is the cable that lets you stop thinking about it. Buy Cat6a for any in-wall run you do not want to pull twice, and keep Cat6 patch leads for the short desk hops if you have them on hand.

The smartest 10 GbE you will buy, though, is not RJ45 at all. The link from the switch to the NAS, if they sit in the same rack or the same closet, should be a passive SFP+ DAC (direct attach copper) cable: a fixed twinax cable with the transceivers molded onto both ends. A passive DAC draws under 0.15 W, has near-zero latency, costs a fraction of two optics plus fiber, and is rated to about 7 meters (checked Jun 2026). For the server-to-switch hop that is the single most important link in the building, DAC is cheaper, cooler, and more reliable than running 10GBASE-T over copper. Reserve fiber and optical transceivers for the rare run longer than 7 meters or between rooms.

So the physical map for the build is short: NAS to switch over a 1 to 3 meter DAC, switch to each heavy desk over Cat6a (or DAC if they are close and you fitted SFP+ NICs), and switch to the everyday desks over whatever multi-gig copper you have. If a desk is across the building, or you are deciding between a wired drop and the wireless you already own, that tradeoff is its own question covered in Wi-Fi versus wired for editing off a NAS.

The full reference build, priced #

Putting the parts together for a four-editor team, network gear only (the NAS itself is a separate decision, covered in build versus buy for a small post house and the 2026 storage appliance roundup). Prices are street figures checked Jun 2026 and will drift, so treat them as the shape of the bill, not a quote.

That is the entire network spine for a four-person room, server NIC aside, for roughly the cost of one editor's Thunderbolt adapter times three. The point is not that it is cheap in absolute terms. The point is where the money lands: two 10G adapters on the stations that need them, one fat 10G link on the NAS, and 2.5G everywhere else. Spend the same budget putting 10GbE on every desk and you have moved the bottleneck to the NAS link you underfunded, which is the most common and most expensive way to get a post network wrong.

The mount layer is what makes 2.5G feel like 10 #

Here is the one place this is genuinely about what we build, flagged as such. A network gives you bandwidth; it does nothing about the chatty protocol overhead that makes a NAS feel sluggish even on a fast link. The reason a single editor is comfortable on 2.5 GbE is that scrubbing reads only the blocks under the playhead, not whole files, and the reason listings and searches feel instant is a local metadata index, not raw throughput. JuiceMount is the mount layer we make for exactly that: block-level streaming off the NAS, a local SSD cache so revisits never re-cross the wire, and a local search index. It is open source and $0 per seat, which is why the closing card points there.

The honest "where it does not fit" matters here. JuiceMount does not make your line faster, and it does not change the rule above: the NAS still needs its fat 10G link, because first-touch reads pay the wire in full for every editor, and the cache only wins on what it already holds. A good mount makes a modest network feel generous; it does not let you skip building the network. If you want the wider framing of why this layer exists alongside the SaaS suites, that is what a real mount means for editors.