We are using for implementing multi tenancy as ***separate-schema per tenant approach*** in our application using Spring data JPA with Hibernate

https://docs.jboss.org/hibernate/orm/4. ... ml#d5e4866
(note: This is also explained as Strategy 3 at following stackoverflow question link --- http://stackoverflow.com/questions/2122 ... te-and-c3p)

The problem is that while executing any query hibernate sets search path by executing a sql and same is done when connection is released. This is causing 2 extra sql execution and database round trips. While evaluating performance we found that the schema setting part is taking almost 30% of the total execution time. Is there any way we can avoid the extra sql executions for setting search path. Please note that we don't have an option to use separate connection pool for a schema

You basically have two options for multitenancy:

1. Either you separate the database or schema at the connection pool level,

2. Or you use a single connection pool and switch the schema at runtime.

According to this StackOverflow answer, the PostgreSQL SET search_path operation should be very fast and cheap. If you experience problems with the PostgreSQL schema setting, you should address this problem on the PostgreSQL forum. Maybe it was some issue that got fixed in a later version.

Vladimir Sitnikov suggested that SET search_path invalidates the Statement cache, which might explain your performance degradation issue.