Multi-tenancy means that many customers share one building, one server or one software platform, while none of them can reach the others' machines or data. Each customer is called a tenant, like the families who rent flats in one apartment building. The owner runs the shared parts once, and every tenant pays a share of the cost.
Here is how it works in general. One provider builds and runs the shared resource, such as a data-centre hall, a cloud server or a software application. Each tenant gets its own separate space inside it. The separation is enforced in different ways: by locked cages in a building, by separate virtual machines on a server, or by a tenant label on every record in a software database, which the software checks on every request. The provider runs the shared parts for everyone: the power, the cooling, the software updates and the security. It measures how much each tenant uses, and bills each one.
The benefit is cost. A small company can use a data centre, a cloud or a large application that it could never afford to build alone. The risk is a flaw in the separation. If a bug lets one tenant read another tenant's data, many tenants are exposed at once. A smaller problem is the noisy neighbour: one tenant using so much of a shared server that the other tenants' work slows down.
Take, for example, a colocation data centre, where a bank, a streaming company and a hospital rent locked cages in the same hall. Guards, cameras and the cages keep each company's servers out of the others' reach.
Consider a shared web hosting plan. Hundreds of small websites, such as a dental clinic's and a coaching centre's, sit on one server to split its cost.
Say several AI start-ups each need GPUs for a few hours. A GPU-as-a-service platform decides whose job runs on which GPU, keeps each customer's work separate, and measures each customer's use.
Mobile networks share in the same way. A distributed antenna system in a mall is one set of antennas that every operator uses.
How is multi-tenancy different from single-tenancy? In a single-tenant setup, the customer gets a server, a cage or a copy of the software of its own. This costs more, but no neighbour can affect it. Most SaaS applications are multi-tenant. Who secures which part of a shared cloud is set out in the shared responsibility model.
Appears in
- Distributed antenna system (DAS)↗ Telecom networksA distributed antenna system in a mall carries the signals of several operators on the same antennas.
- Colocation↗ Data centresA colocation data centre houses many companies' servers in one building, each locked in its own cage.
- Cross-connects and interconnection↗ Data centresAn interconnection data centre houses many tenants, and each one connects only to the networks it chooses.
- Web hosting↗ Cloud hostingShared web hosting runs many small websites on one server, and each site cannot see the others' files.
- Private cloud software↗ Infrastructure softwarePrivate cloud software lets an organisation rent its servers to several departments or customers, keeping each one's data apart.
- GPU-as-a-service platform↗ Infrastructure softwareA GPU-as-a-service platform shares one room of GPUs among many customers, and no customer can see another's data.