Tenant offloading is the practice of moving an inactive tenant’s data out of active memory and into cheaper cold storage, then reactivating it on demand when the tenant becomes active again. It is part of managing a tenant lifecycle in a multi-tenant system, where tenants pass through states such as hot, cold, and frozen.
The motivation is cost efficiency at scale. In a system with many tenants, only a fraction are typically active at any moment, yet keeping every tenant’s vectors in fast memory is expensive. Offloading idle tenants to object storage frees up costly resources, and the tenant’s data is restored to a hot, queryable state when needed — accepting a brief reactivation delay in exchange for substantial savings.
This capability is closely tied to data tiering and is valued in SaaS applications with large, uneven tenant populations, such as those with many trial or dormant accounts. A formal tenant lifecycle that supports offloading and reactivation lets a provider serve a vast number of tenants economically, paying for fast storage only on the tenants that are actually using the system.