Home Page for the TeradataForum
 

Archives of the TeradataForum

Message Posted: Tue, 13 Aug 2003 @ 00:44:02 GMT


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


Subj:   Re: PSF questions
 
From:   Greene, Thomas L

I think you have the right general idea, but let me offer some things to consider.

1) Moving users out of the default partition is (in my humble opinion) a good idea. I only run administrative type work in RP0.

2) Since weights are relative, and based only on active performance groups, why not make RP0 significantly higher than the other resource partitions.

My system is set up so that RP0 equals the weight of the other 4 partitions combined.

rp0=1000
rp1=400
rp2=300
rp3=200
rp4=100

Users running in rp1 can still use 100% of the system, but my admin (me) can run faster when he gets on the system.

3) Limits are useful, but my testing shows that they can be set too low. I certainly wouldn't go much below 75%.

4) Why set the batch priority so high? If you expect the resource usage of normal ("general") users will be lighter during night time hours, then you might also expect that batch users will get more resources regardless of their weight. Keep in mind that performance periods cannot cross resource partition boundaries. Thus, the more active performance groups you have, the less difference in relative weight you'll be able to achieve with performance periods. Also, the lower the relative weight of a resource partition with respect to other resource partitions, the less effect you'll be able to get from performance periods within that resource partition.

5) Performance periods don't cause the user to run under a different account. Users are associated with accounts, and accounts are associated with performance groups. Performance periods associate different allocation groups, each with an assigned weight, to a performance groups based on time (for example). In your case, the users' accounts won't change. Nor will their performance group. Only the allocation group, and hence their relative weight, will change.

6) I consider it a good practice to weight things in order. For instance, rp0 should be weighted higher than rp1, which is weighted higher than rp2, etc. The same is true for weighting the allocation groups. Consider this...

schmon -a 80 N DEF 25
schmon -a 81 N DEF 100
schmon -a 82 N DEF 1
schmon -a 83 N DEF 1

...can be changed to ...

schmon -a 80 N DEF 1
schmon -a 81 N DEF 25
schmon -a 82 N DEF 100
schmon -a 83 N DEF 150

In either case, if only one performance group is active within a resource partition, it will get 100% of the weight associated with that resource partition. Thus, it will have the same priority (relative weight) regardless of what time of day it is.

7) If you don't need all five resource partitions, consider placing the users and batch in the same resource partition. That way you can use performance periods more effectively to contrast the priorities of the two workloads.

8) Your syntax for defining performance periods looks correct to me. If you're running V2R5.0, consider using the Priority Scheduler Administrator (PSA) tool. It makes setting up your system easier. You don't need to figure out the syntax, but if you're interested, PSA will show you what the syntax will be.

Just food for thought!



     
  <Prev Next>   <<First <Prev Next> Last>>  
 
 
 
 
 
 
 
 
  
  Top Home Privacy Feedback  
 
 
Copyright for the TeradataForum (TDATA-L), Manta BlueSky    
Copyright 2016 - All Rights Reserved    
Last Modified: 15 Jun 2023