I have a huge table (hundreds of millions rows) with this structure: call_id, date, cust_id. The PK is cust_id + call_id, so call_id is unique for each cust_id.
I need to update call_id to be unique in whole table and sequential when table is ordered by cust_id and date. Values can come from sequences or generated by code.
Of course, performance is very important.

Any ideas?

Thanks,
Radu