Home Page for the TeradataForum
 
https:

Archives of the TeradataForum

Message Posted: Thu, 12 Apr 2007 @ 11:39:15 GMT


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


Subj:   Re: Merging large volumes of data into large history tables
 
From:   Walter, Todd A

A couple clarifications:

2. Teradata does not run at the speed of the slowest hardware node. This is a commonly repeated misconception but it is not true. Teradata provides "co-existence" wherein we configure differing numbers of AMPs on different performing nodes such that each node has a proportion of the work to match it's proportion of the total system power. This allows older hardware and newer hardware to participate in the same configuration without holding the entire system back to the performance of the slowest node.

4. There currently is not a plan to provide "no-logging" for permanent tables. The logging is critical not only for transaction consistency but also for structural integrity of the table. Recent releases have significantly improved the performance of the methods used for journaling and the journaling itself. Plus we have radically improved the rollback logic such that the rollback will now only take the length of time of the original operation.

3,5: Certainly true that newer hardware runs faster on a per node basis and that newer node technologies deliver very impressive performance. But all the things we have been discussing on this thread are software improvements in recent releases which apply on any platform level, not just on the latest. And of course Teradata does not rely on node level performance but on the ability to scale to the number of nodes (of any generation) to solve a problem - just takes less nodes of the newer generation to solve the same problem at the same performance.



     
  <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: 28 Jun 2020