I have the application that has multiple threads reading and writing files from/to same Centera pool.

What is the ‘preferred’ pool connection model in this case?

1. One pool reference is shared among all threads
   
2. Each thread manages its own pool reference.
   

Is there any advantage in performance or resource usage between these two approaches?

I am currently using the first model, but I see drawbacks in error handling – e.g. when the pool connection is lost the recovery steps need synchronization.