UK How direct attached storage works
Direct attached storage (DAS) is any storage connected directly to a single computer, with no network between them. The drives appear to the operating system as local disks, and the computer reads and writes them at block level, the same way it uses its boot drive.
DAS covers more than people expect. The SSD inside a laptop is DAS. So are the hot-swap drive bays on a server, an external RAID enclosure cabled to a host over SAS, and a USB or Thunderbolt drive on a desk. What they share is a one-to-one link: one host owns the storage, and other computers can reach it only through that host.
Because nothing sits between the processor and the drives except a controller and a cable, DAS has the lowest latency and the simplest setup of any storage architecture. It also keeps working when the network is down, which matters a great deal at the edge.
DAS vs NAS vs SAN
DAS is one of three storage architectures. The other two put storage on a network so many computers can share it, and they differ in what the network carries.
| Feature | DAS | NAS | SAN |
|---|---|---|---|
| What it presents | Local disks (block) | Shared folders (file) | Network disks (block) |
| Connection | SATA, SAS, NVMe, USB, Thunderbolt | Ethernet | Fibre Channel or Ethernet |
| Protocols | Drive protocols, no network stack | SMB, NFS | Fibre Channel, iSCSI, NVMe-oF |
| Shared between hosts | No, one host | Yes, by many clients | Yes, volumes assigned to hosts |
| Latency | Lowest | Highest | Low, depends on the fabric |
| Typical use | Edge systems, servers, workstations | File sharing, backups, media libraries | Virtualisation clusters, databases |
NAS and SAN both depend on network bandwidth. On a site or vehicle network, that often means 10GbE networking or faster before shared storage can match a local NVMe drive.
What a storage array is
A storage array is an enclosure of many drives managed as one system. The simplest kind, JBOD (just a bunch of disks), passes each drive through to the host. A RAID array adds controllers that combine the drives into protected volumes, with a cache, spare drives and tools for replacing failed disks while the system runs.
An array can be DAS or SAN, depending on how it connects. Cabled over SAS to one server, it is DAS. Connected to a Fibre Channel or iSCSI network and shared between servers, it is a SAN array. The drives inside may be the same in both cases; only the connection and the way volumes are presented change. In smaller systems the array is often built into the computer itself, as RAID across its internal bays, as on many rack mount servers.
Why edge systems use DAS
At the edge, DAS is usually the only sensible choice. Systems sit on vehicles, trackside or in remote cabinets where a network may be slow, intermittent or absent, and the data has to be captured locally first.
-
//
Recording that cannot depend on the network. Industrial NVRs write every camera stream to internal drives, often in RAID, and send clips or alerts upstream when bandwidth allows.
-
//
Data offload from vehicles. Many in-vehicle computers have removable SSD trays, so a test vehicle or train can hand over terabytes of data by swapping a drive instead of uploading it.
-
//
Low-latency AI pipelines. Local NVMe drives feed models and store results without the delay of a network round trip.
-
//
Edge storage nodes. Multi-bay edge servers with NVMe or SATA bays act as local DAS for their own applications and can share data to the site over the network when needed.
Limits of DAS and how to work around them
The one-to-one link that makes DAS fast and simple is also its main limit. Capacity is tied to one host, so spare space on one machine cannot help another, and if the host fails its storage is offline until the host is repaired or the drives are moved. Growth is capped by the number of bays and ports on the computer or enclosure.
Each limit has a standard answer. RAID across internal bays protects against a drive failure, and hot-swap trays let a failed drive be replaced without shutting the system down. Scheduled replication to a central NAS or SAN covers loss of the whole host, and only needs bandwidth outside working hours. Choosing a chassis with spare bays, or a free PCIe slot for an NVMe adapter or SAS controller, leaves room to grow. Treat the edge copy as the working copy and the central copy as the backup, and DAS covers most edge workloads well.
Quick answers on direct attached storage
Direct attached storage: common questions
What is DAS storage?
Storage wired directly to one computer. Internal drives, drive bays and external enclosures on SATA, SAS, NVMe, USB or Thunderbolt are all DAS, because no network sits between host and drives.
Is an external USB drive DAS?
Yes. A USB or Thunderbolt drive is the most common form of external DAS. Multi-bay enclosures add RAID, but they are still DAS while they connect to a single computer.
What is the difference between DAS and NAS?
Where the file system lives. DAS gives one computer raw disks to format and manage. A NAS runs its own file system and shares folders with many clients over Ethernet.
Can DAS be shared between computers?
Not directly. The host can share DAS over the network like any local drive, and some SAS enclosures connect to two servers for failover, but DAS is built for one owner.
Is DAS faster than NAS?
Usually, yes. With no network or file-sharing protocol in the path, DAS has lower latency, and a local NVMe drive outruns most network links. A NAS on fast Ethernet can come close for large sequential transfers.
We fit the drives, bays and RAID configuration to every edge computer and server we build as standard, sized for the recording, logging or AI workload and the conditions it will run in.