Concurrency Control
When not running an Online Archive, the ARCHIVE statement places a HUT read lock on the:
If there is an existing SQL read lock or access lock (created by using the LOCKING FOR ACCESS modifier) on the database, Teradata ARC can obtain its read lock and start the archive. If there is an existing SQL write lock or an exclusive lock on the database or table, Teradata ARC will be blocked until the blocking lock is released. Once the read or access lock is obtained, it is maintained on that object until it's data is archived and then the lock will be released.
When running an Online Archive, the ARCHIVE statement places a transaction table read lock on all of the object(s) to be archived to establish a consistency point so that Online Logging can be started on all of the involved objects at the same time. After the consistency point is established and Online Logging is enabled, the read lock is released prior to archiving any of the table data.
Establishing the consistency points defines the state of the tables to which they will be restored. While waiting for the table read locks required to set the consistency point, new write transactions will wait behind those read lock requests. If it is important to reduce impact to the write transactions, then there are some important considerations for reducing the time to establish a consistency point. New write transactions can be expected to wait for at least as long as the longest running write transaction in the system that accesses one of the tables in the consistency point. The options to shorten the wait time on the new write transactions are four fold.
Introduction to Teradata contains details on concurrency control and transaction recovery.