http://www.oracle.com/technetwork/database/oracledrcp11g-1-133381.pdf
Master Note: Overview of Database Resident Connection Pooling (DRCP) (Doc ID 1501987.1)
Is Database Resident Connection Pooling (DRCP) Supported with JDBC-THIN / JDBC-OCI ? (Doc ID 1087381.1)
How To Setup and Trace Database Resident Connection Pooling (DRCP) (Doc ID 567854.1)
How to tune Database Resident Connection Pooling(DRCP) for scalability (Doc ID 1391004.1)
Connecting to an already started session (Doc ID 1524070.1)

Managing Processes http://docs.oracle.com/cd/E11882_01/server.112/e25494/manproc.htm#ADMIN11000<— HOWTO
Example 9-6 Database Resident Connection Pooling Application http://docs.oracle.com/cd/E11882_01/appdev.112/e10646/oci09adv.htm#LNOCI18203
When to Use Connection Pooling, Session Pooling, or Neither http://docs.oracle.com/cd/E11882_01/appdev.112/e10646/oci09adv.htm#LNOCI16652
Database Resident Connection Pooling and LOGON/LOGOFF Triggers http://docs.oracle.com/cd/E11882_01/server.112/e25494/manproc.htm#ADMIN13400
Example 9-7 Connect String to Use for a Deployment in Dedicated Server Mode with DRCP Not Enabled http://docs.oracle.com/cd/E11882_01/appdev.112/e10646/oci09adv.htm#LNOCI18204

DRCP with JDBC

supported starting 12c
12c New Features http://docs.oracle.com/database/121/NEWFT/chapter12101.htm#NEWFT182

my thoughts on this

On one of our database here, there are multiple app schemas (around 6) tied to a JVM program and each of them has their own connection pool setting to around 250 max. The problem with this app setup is you don’t have a “central pool” of connection pools. And there will be a time where one schema is overloading the box vs the others. The issue here is that if we bring the pool down to a lower value (30-50 min/max) to let the app users queue for available pool on the app side and not on the DB side we have to do it across the board on all schemas because if we do it on just one/two apps they would still suffer as the database load is not reduced due to other applications still loading the database. Now lowering the value for everyone would make it faster but with so many development teams involved it is politically not easy to convince especially if it’s a bank. So unless you have a parallel environment (clone) where you can mimic the load with the proposed changes then you may see some progress on this effort. This one is really a tricky issue to control.

We explored a couple of things:

1) DRCP was a promising option for us but it is only supported in 12c (12c New Features http://docs.oracle.com/database/121/NEWFT/chapter12101.htm#NEWFT182) but there are limitations to this feature as well like DRCP only has one (default) pool, limits to this feature like ASO, etc.
2) Connection Rate Limiter on the Listener side, there’s a white paper http://www.oracle.com/technetwork/database/enterprise-edition/oraclenetservices-connectionratelim-133050.pdf and it seems like it just queues the sessions and doesn't kill them. You can see the demo here http://jhdba.wordpress.com/2010/09/02/using-the-connection_rate-parameter-to-stop-dos-attacks/