Archives of the TeradataForum

Message Posted: Wed, 12 Jan 2005 @ 14:25:04 GMT

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

Akleema:

You are certainly welcome for my speculation on the meaning of these values.

As for the CPUs themselves, at some point over the course of running Teradata all of the CPUs will most likely at some point run AMP tasks and all of the CPUs will have run PE tasks. NCR nodes use all CPUs equally based on current requirements. All the CPUs have access to all the same resources of memory, system buses, etc. So instead of nailing a task to a CPU and making it wait if another task is using that CPU when the nailed task needs CPU time again, the system simply assigns a different CPU to do the work for the task.

There is no "manager" CPU and "worker" arrangement for the use of the CPUs. They all do the work needed by any and all tasks as the task needs them and they are available.

My understanding is that as any AMP uses CPU time, that time goes into an accumulator bucket. As any PE uses CPU time that goes into a different accumulator bucket. I believe what you are seeing is the total accumulation in each of these buckets for all of these two types of tasks. An AMP consists of somewhere around 90 UNIX tasks and a PE is comprised of around 60 UNIX tasks.

I hope this gives you a better idea of my take on these usage values. If any one has a more precise explanation, I am sure we will see it out here shortly.

Regards,

Michael Larkins
Certified Teradata Master
Certified Teradata SQL Instructor

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