Archives of the TeradataForum

Message Posted: Fri, 07 Dec 2001 @ 11:13:27 GMT

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

There are following 3 Workload management techniques which can be used:

Limiting spool:
The limiting of spool places a cap on the amount of temporary disk resources a process can consume before aborting. Limiting spool can help limit the consumption of non-optimal queries, but it cannot do anything to mitigate query performance before it passes the spool threshold.

Production schedules:
Even the most unstructured data warehouse uses some type of production schedule. Though the idea of a production schedule might seem painfully obvious, it is important to consider the concept of performance periods that is, what type of work you are required to complete within a given time/day/week. Conversely, you must also consider what type of work should not occur within the same performance period.

PSF settings:
A session always runs under a performance group. A process in Teradata is allocated more or less access to system resources according to the:

--- Performance group of the session or request

--- Resource partition(s) you define for each performance group

--- Scheduling policy you establish for the partition’s performance period

Performance groups/allocation groups: Although a resource partition is the major division of system resources, Teradata provides the ability to define a performance group/allocation group combination to assign and prioritize work within a partition.

On UNIX, you can prefix an account string with a performance group code. The code requires a leading dollar sign ($); for example, ’$LPUBS’.

Note: The weights and schedules you define affect all AMPs, but are scheduled within each AMP according to workload.

STEP 1
IDENTIFY Resource Partition

Uniquely identify this partition
Associate this partition with a user account

You can include a performance group name in the account string prefix. (You must have already created the name and partition using the schmon utility of the Priority Scheduler Facility.) Bracket the name with leading and trailing dollar sign ($) characters; for example, ’$L$RESPARTNAME$PUBS’.

COMPUTE WEIGHT
compute the percentage of resources to be assigned to this partition. This is a relative weight; to find the percent value, the weight of each partition is divided by the sum of the weights of all active partitions.

For example, if three active partitions have weights of 10, 20, and 30, respectively, then each would receive the following percentage of available system resources:

10 = 17% (10/60)
20 = 33% (20/60)
30 = 50% (30/60).

STEP 2

Performance group
subdivide a partition’s resource allocation according to the defined priority level.

Performance associate this performance group with

L   0      Low       One half of computing resources of M
M   2      Medium    Defaults
H   4      High      Twice the computing resources of default
R   6      Rush      Four times

Associate this performance group with an allocation group.

The weight can be allocated in one of the following ways:

Default Allocation – proportional group weights

Immediate - Similar to Default, except all tasks run equally within a group

Absolute - An absolute percent value (n%). “ not to exceed n% of CPU” limit.

Relative - Similar to Absolute, policy, except weight is relative to elapsed time and workload.

Use the schmon and xschmon utilities of the Priority Scheduler Facility to establish the resource partitions, scheduling policy, milestone limits, and allocation groups that suit your workloads.

Regards

ROHIT REVO

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