Managing Shared Resources in a Distributed Computing System

A distributed computing system includes several partitions that each contain a separate copy of shared resources that receive modifications via behaviors and transactions specified by a user. The transaction manager performs the requested behavior or transaction in parallel on each copy of the shared resources as indicated by a resource ID. This allows the distributed computing system to operate in parallel without competing for the same shared resource, avoiding deadlocks and race conditions. If a behavior or transaction fails while modifying a copy of a shared resource, the transaction manager prevents the behavior or transaction from modifying the remaining copies and preempts results from the failed behavior or transaction. The transaction manager reestablishes a consistent state across shared resources by rolling back the failed behavior or transaction, reverting each copy of the shared resources to its state prior to executing the behavior or transaction.