Home Page for the TeradataForum
 

Archives of the TeradataForum

Message Posted: Mon, 14 Nov 2005 @ 15:13:36 GMT


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


Subj:   Re: Scalability in Teradata - in terms of AMPs
 
From:   Morris, Mark

In general, the answer is yes, but as you allude to with your question on performance there are other considerations.

Your NCR account team can help you explore the options through the use of configuration and performance modeling tools.

Some of the needed information is:

1. Do you need more storage capacity?

2. Do you have a surplus of CPU cycles?

3. Are you bringing a new application online? Growing the data on existing applications? Trying to more efficiently use the CPU for existing applications?

4. Are there available disk slots in your disk arrays, or alternately could you add a new disk array to provide the storage for the new AMPS?


Your performance question can't be answered without going through analysis and modeling.

For example, I will give two hypothetical situations:

A. The nodes have 8 amps each. The nodes are underutilized (peak at 80%) because the application tends to be IO intense. There are available slots to add disks to suport 2 more amps per node. The disks are added, 2 amps are added. A reconfig is executed to move data to the new AMPS. Afterwards the CPU utilization peaks at 100%. Immediately after the change (prior to adding new applications or more data) the performance of the system increases by 25%. It is actually processing more rows per second. However, as more data is added to the application, the performance in queries per hour will return to previous levels.

B. The nodes have 8 amps each. The nodes are 100% utilized during peak periods. As in example A, 2 amps and their disks are added and a reconfig is done to give data to the new amps. Immediately after the change the performance of the system is basically the same, since the limiting factor was the CPU and it is still the limiting factor. However, now lets say the application adds new data. Then the queries may take longer depending on the access path. In terms of queries per hour, the perceived performance by the business user may decrease. Even though the system is still processing the same number of rows per second, there are more rows.


There is a lot that goes into capacity planning and this is only the tip of the iceberg.


Mark



     
  <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