Bulkhead Pattern

Bulkhead: named after ship compartments that limit flooding to one section. In software, it partitions services, threads, or connections so degraded or slow dependencies cannot monopolize shared resources, protecting unrelated flows.

Thread pool bulkhead: each external service gets its own bounded thread pool. If service A becomes slow, its pool fills and calls queue or fail, but service B's pool is untouched, so callers of B continue to get timely responses.

Semaphore bulkhead: uses a counting semaphore to cap concurrent calls; the caller thread itself executes the call once a permit is acquired. Thread pool bulkhead uses a separate pool, providing true thread isolation and timeout support.

Microservices relevance: in a mesh of many services sharing infrastructure, a slow payment service could exhaust a shared thread pool, making even unrelated catalog queries fail. Bulkheads provide isolation walls between service integrations.

Complementary resilience: bulkheads contain the blast radius by limiting resources; circuit breakers stop futile calls after failures are detected. Used together, they give comprehensive protection: limit concurrency AND fail fast when a service is down.