Home Page for the TeradataForum
 

Archives of the TeradataForum

Message Posted: Fri, 09 Sep 2005 @ 16:01:24 GMT


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


Subj:   Re: Client Request Load Balancing
 
From:   Diehl, Robert

Clay,

The link Victor sent talks about how it decides which Ethernet connection on the nodes to choose. This part is used to spread the Network Traffic around so you shouldn't get a network side bottleneck. After it reaches the nodes Teradata then sends the session to a Parsing Engine (PE). Somehow it chooses a Parsing Engine with lower traffic. I do not know the details on how this is decided. The Parsing Engine then does its work to send the requests to the VPROCS that are needed to complete the request.

The Parsing Engine decision becomes very important for Active Data Warehousing. Since you want Tactical Query requests to be cached for optimal sub-second responses, you don't want to spread the requests among to many PE's. Reducing the number of PE's used for tactical queries can be accomplished in two ways. Both techniques might be needed.

1) Use Session Pooling to reduce the number of sessions. Since plans are cached at the Parsing Engine level, having too few requests on each PE will cause fewer requests to use cached plans. If you have too many sessions the requests will be spread among many different parsing engines.

2) Use a different Host name for Tactical Queries. Then only assign as many Nodes/NICs/PE's to that host as needed for volume of requests and availability. This works because the Parsing Engine routing is done only amount PE's on the same host. This also allows you to only allow specific users to use certain hosts for security (Revoke Logon). It also prevents PE bottlenecks from hurting Tactical Query Performance.


Hope this explains it in enough detail.


Thanks,

Bob Diehl



     
  <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