New data will inserted every day so, we would want to keep only 2 weeks worth of data. Hence we would want to implement something recursive which will delete on daily or weekly basis. (Without hurting the logs)
I'd recommend listening to what Luis and Sean recommended, or some combination. Also, "Your Name Here" make great points in suggesting ways to evaluate the "weekly" constraint and the granularity of the data.
Finally, some combination of the above, along with possibly using a backup to do a lot of the purging off of the production system and then adding the difference (whatever was added since the backup) might alleviate stress on the production log files.
It also sounds to me like Full Recovery with log backups should be planned for, but the experts here know better than I do whether that is feasible for a db with 2 billion rows (might make sense after the purge).
"I love spending twice as long and working twice as hard to get half as much done!" – Nobody ever.
"Operator! Give me the number for 911!" - Homer Simpson
"A SQL query walks into a bar and sees two tables. He walks up to them and says 'Can I join you?'"