This tactic is aimed at reducing the impact of sudden spikes in your system's load. Flooding a service with requests can cause performance issues, reliability ones, and even dropping valid requests. Once again, queues are there to save the day.
To implement this pattern, we just need to introduce a queue for the incoming requests to be added asynchronously. You can use Amazon's SQS, Azure's Service Bus, Apache Kafka, ZeroMQ, or other queues to achieve that.
Now, instead of having spikes in incoming requests, the load will get averaged. Our service can grab the requests from the said queue and process them without even knowing that the load was increased. Simple as that.
If your queue is performant and your tasks can be parallelized, a side benefit of this pattern would be better scalability.
Also, if your service isn't available, the requests will still get added into the queue for said service to process when it recovers, so this may be a...
