Archives of the TeradataForum

Message Posted: Tue, 25 Apr 2007 @ 01:53:52 GMT

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

A starting point is to understand the bottleneck which might be Teradata capacity/contention, client cpu capacity/contention, client network or i/o. For client cpu, multiload output reports the client cpu which you can compare to acquisition phase elapsed time. If they are close, it is a client cpu bottleneck. Multiload client cpu can be high with vartext, nulliff, apply where. It uses much less client cpu with binary/fastload format data and no field manipulations since it can copy the rows directly rather than have to go field by field. Also, 1 session is definitely a bottleneck. With multiple sessions (8 is usually good) Multiload should never have to wait for Teradata but will always have another session on which it can send the next buffer.

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