In most systems, we use remote calls. Many factors may have an impact on these remote calls e.g. network latency, server availability and so on. So we should assume that something can go wrong. These calls can be potential bottlenecks, we don’t want user waiting for the response from the server very long, because external API is very slow or not available. Also if we have a few services which communicate with each other we shouldn’t aggravate the situation when one of them has too much traffic and slow down significantly. So how to do it correctly?