Perinstance

Per-instance, often hyphenated as per-instance, is a descriptive term used in software engineering to indicate actions, configurations, or data that apply to each individual instance of a component, service, or object, rather than to a system-wide or global scope. This distinction helps distinguish what is tied to a single running copy from what is shared across all copies.

In object-oriented programming, per-instance state is stored in instance fields that belong to each object independently, contrasted with per-class or static state that is shared across all instances. This affects lifetime, encapsulation, and thread safety, since per-instance data does not inherently interfere with other instances.

In deployment and cloud environments, per-instance configuration means each running copy of a service (an instance) can have its own settings, environment variables, or resource limits. Per-instance pricing and allocation describe billing and provisioning that scale with the number of active instances, such as virtual machines or containers, enabling fine-grained control but increasing management overhead.

In observability and telemetry, per-instance metrics track data separately for each instance, enabling isolation and fault diagnosis when some instances diverge from others. In testing and development, per-instance fixtures or mocks provide isolation between test cases by running objects in their own contexts.

Trade-offs of per-instance design include greater customization and fault isolation at the cost of increased resource usage and management complexity. It often requires additional configuration management and coordination to ensure consistency across multiple instances.

Related concepts include per-class, per-tenant, instance isolation, and scalable architectures.