Another way you might be able to approach hash-based distribution among nodes is using GridSQL.
Instead of putting the logic for what node to execute against directly into the database, GridSQL provides a client that speaks to a node coordinator process that handles that job. This process is all relatively transparent to you. The main catch is that not all PostgreSQL features are available, and you need a client driver compatible with GridSQL.
GridSQL uses its own parser, planner, and optimizer, accepting American National Standard Institute (ANSI) standard SQL, but not all of PostgreSQL's syntax. The driver interface for your applications to talk to the database can look like a standard PostgreSQL one, but not everything ...
