Storage fundamentals
JBOD
JBOD stands for "just a bunch of disks." It describes drives that are presented to the operating system individually, with no RAID controller combining them or adding redundancy. The term is also used for an external enclosure that holds many such drives and connects them to a server.
How does JBOD differ from RAID?
The difference is where redundancy is handled.
| Aspect | JBOD | RAID |
|---|---|---|
| How drives appear | Each drive individually | One or more combined volumes |
| Redundancy | None at this level | Mirroring or parity within the group |
| Controller | Simple host bus adapter | RAID controller or software RAID layer |
| After a drive failure | That drive's data is unavailable locally | Group continues, rebuild onto a spare |
| Usable capacity | All of it | Reduced by mirroring or parity |
Why use disks with no redundancy?
Because the redundancy is provided somewhere better placed to do it. Distributed and software-defined storage systems protect data across servers using replication or erasure coding. That protects against the loss of a whole server or rack, which RAID inside one server cannot. Adding RAID underneath would spend capacity twice, hide individual drive health from the storage software and tie rebuilds to a single group of drives. Presenting plain disks lets the software manage placement, monitor each drive and rebuild across the cluster.
What is a JBOD enclosure?
A JBOD enclosure, or expansion shelf, is a chassis containing drive bays, power supplies, fans and connection hardware, with no processors or storage intelligence of its own. It attaches to a server through SAS cables and can add dozens of drives, sometimes more than a hundred, to a single storage node. Enclosures raise density and lower cost per terabyte. They also increase the amount of data that depends on one server, which the cluster's protection scheme and node count have to account for.
What should be considered when using JBOD?
Three points matter. The software above must provide the redundancy, since JBOD offers none. Drive failures are handled by the storage software, so its monitoring has to see each device directly, which is why controllers are set to pass drives through unmodified. And failure domains change: a dense enclosure that loses its connection takes many drives offline at once, so placement rules should treat the server and its enclosures as one unit. Object storage sizing covers how drive counts per node feed into the design.
Frequently asked questions
Is JBOD a RAID level?
No. It is the absence of RAID. Some controllers offer a JBOD or pass-through mode alongside RAID levels.
Is data on JBOD unprotected?
At the disk level, yes. Protection has to come from the storage software or from copies elsewhere.
Why not use RAID under a distributed system as well?
It duplicates protection the system already provides, reduces usable capacity and hides drive-level information the software uses.
Related terms
- RAID: redundancy within a group of drives.
- Software-defined storage: software that manages plain disks directly.
- Storage node: the server a JBOD enclosure attaches to.
- Erasure coding: the redundancy used above JBOD.
- Annualized failure rate (AFR): how often individual drives fail.