Archives of the TeradataForum

Message Posted: Tue, 14 Aug 2012 @ 23:39:14 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

If no AWTs are available, work messages that arrive from the dispatcher on the AMPs are temporarily held in a message queue. Each AMP has one message queue, which, unlike the query delay queue, does not function on a first-in, first-out basis. Consequently, messages waiting in the queue are not treated equally.When multiple messages are waiting, the request that gets the next AWT will depend on where the message sits on the queue. A message represents one step in a query request. The ordering on the queue is by:

1) Work type, in reverse sequence (WorkNew is always last)

2) Priority scheduler relative weight within each work type

This means that messages representing low-priority work will wait longer for an AWT than higher-priority work messages, and work that is starting a new step (WorkNew work type) may wait longer than messages of the same priority that represent spawned work (WorkOne work type). The ordering was designed to give work that has already begun and work that is more important an advantage in acquiring AWTs. So it is possible that when there is a shortage of AMPs, some messages will wait longer for an AWT

Thank You,

Geeta Kudumula
TERADATA Certified Master

	<Prev	Next>		<<First	<Prev	Next>	Last>>