When using the new OT trace database in a multi-processing setting (e.g., CECA.py), the new DB is slower than the older CW DB. The reason for that is how the database connection with SQLAlchemy is established. Instead of opening a new connection each time, we should keep the connection open and allow different Ray workers accessing them.
When using the new OT trace database in a multi-processing setting (e.g., CECA.py), the new DB is slower than the older CW DB. The reason for that is how the database connection with SQLAlchemy is established. Instead of opening a new connection each time, we should keep the connection open and allow different Ray workers accessing them.