Selective scaling is related to the previously discussed polyglot architecture. In this context, all functional modules may not require the same level of scalability. Sometimes, it may be appropriate to determine boundaries based on scalability requirements.
For example, in the hotel booking scenario, the Search microservice has to scale considerably more than many of the other services, such as the Booking microservice or the Notification microservice, due to the higher velocity of search requests. In this case, a separate Search microservice could run on top of an Elasticsearch, or in a memory data grid for better response.