A Map of the Patterns

Protect yourself, protect others

Resilience patterns fall into a few families. Patterns that protect the caller from bad dependencies: timeouts (stop waiting), circuit breakers (stop calling a failing dependency), bulkheads (limit how much of your capacity one dependency can consume) and fallbacks (serve something useful instead of an error). Patterns that protect the callee from too much demand: rate limiting (cap requests per client), load shedding (reject excess work early), concurrency limits and back-pressure (signal callers to slow down). Patterns that handle transient faults: retries with backoff and jitter, idempotency so retries are safe, and hedging for tail latency. They are applied in client libraries (Resilience4j, Polly), in gateways and service meshes (Envoy, Istio), and in infrastructure (load balancers, autoscalers). The rest of this course covers each family and how they combine.

Patterns at a glance

Who each pattern protects and where it usually lives.

pattern               protects   typical location            main parameter
--------------------  ---------  --------------------------  ---------------------------
timeout               caller     every client call           duration (from latency p99)
retry + backoff       caller     client / mesh               attempts, backoff, jitter
circuit breaker       both       client / mesh               failure rate, open duration
bulkhead              caller     client (pools, semaphores)  max concurrent calls
fallback              user       client                      what to return instead
rate limiter          callee     gateway / server            requests per window per key
load shedding         callee     server / LB                 queue time or concurrency
concurrency limit     callee     server                      adaptive in-flight limit
idempotency key       data       server                      key retention window

Safety systems in a car

Brakes (timeouts), anti-lock braking (backoff), airbags (fallbacks), crumple zones (bulkheads) and speed limiters (rate limits) each handle a different danger. You want all of them, configured for your car, not just one big airbag.

त्वरित जाँच: Which pattern primarily protects a service from receiving more requests than it can handle?

• Timeout
• Fallback
• Hedged requests
• Rate limiting or load shedding