A RAC cluster always needs shared storage for database files: each cluster node cannot have its own separate storage.
You would need at least a 3rd physical database server for the standby database that can work without Data Guard provided that you either use you own scripts to send and apply archived redo logs or that you use a product like dbvisit.
I don't think it is possible to apply redo immediately without Data Guard.
Thank you for your reply.
Our current architecture already uses own-written scripts for applying archived redo logs to the standby node. The logs are applied every half hour.
Failover is a manual action, which involves manually applying the redo logs of the last half hour to the standby node and "promoting" the standby to primary.
I will evaluate dbvisit, it looks like this product is able to handle the Data Guard functionality and can be configured to perform failover automatically.