Bulkheads are used in ships to create separate watertight compartments that serve to limit the effect of failure, ideally preventing the ship from sinking. If water breaks through the hull in one compartment, the bulkheads prevent it from flowing into other compartments, limiting the scope of the failure.
The same concept is useful to limit the scope of failure in the architecture of large systems, where you want to partition your system to decouple dependencies between services. The idea is that one failure should not cause the entire system to fail, as shown in the following diagram:
In the bulkhead pattern, it's better to isolate the element of the application into the pool for service, which has a high dependency; so, if one fails, others continue to serve upstream services. In the preceding diagram, Service 3 is partitioned into two pools from a single service. Here, if Service 3 fails, then the impact of either Service 1...
