Archives of the TeradataForum
Message Posted: Mon, 12 Jan 2004 @ 17:51:01 GMT
Subj: | | Re: Explain step 1 - lock pseudo table |
|
From: | | Anomy Anom |
<-- Anonymously Posted: Monday, January 12, 2004 10:20 -->
A simplified explanation would be:
Since Teradata distributes a table across all amps, it is possible (without the pseudo table mechanism employed) that one process would grab a
lock on a table on one amp and a different process grabs a lock on the same table on a different amp. They would both be trying to grab a full
lock and neither could get it because the other holds out a lock that is needed.
Therefore, in order to avoid this deadlock situation, one amp effectively serves as the starting amp (see the physical implementation class or
the Teradata manuals for information on how) for all locks on the table and this is called a "Pseudo Table lock". Like a semaphore, it prevents
deadlock on the locking of a table. Once a process has a pseudo lock, it can lock the table on all amps.
|