Archives of the TeradataForum

Message Posted: Fri, 06 Feb 2004 @ 21:19:53 GMT

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

Yes each AWT is a Process. They are the tasks that process the data to solve the query.

The Value NPROC is a Kernel Tunable setable at the time you do the IDBUILD. ( idbuild is run automatically from inside ACTIVATE_PDE )

According to the MTUNE file the MAX NPROC can be set to is 10240. You system may vary. you can look in your

grep NPROC /etc/conf/cf.d/stune.current

file to see what your current setting is. This file is READ only and gets rebuilt on every idbuild so do not edit this file.

As part of installing teradata we change NPROC to an appropriate value based upon the number of AMPs and PEs you have configured for a node or CLIQUE taking into account vproc migration.

The current default is 8000 which I beleive will exceed the basic configuration of AMPs and PEs that could migrate over to a node.

On v2r5

Each Amp supports about 100 processes.

each PE supports about 160 Processes.

You can multiply it out for your cliques and see how your nodes fair against this value.

If all the vprocs in a 4 node clique migrated over to a single node.

for example --

      56 amps (5600)
     + 8 PE's (1280)
 per clique == 6880 processes

8000 is more than enough to handle the failover and allow additional room for the BASE unix processes.

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