Home Page for the TeradataForum
 
https:

Archives of the TeradataForum

Message Posted: Fri, 26 Feb 2016 @ 20:52:57 GMT


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


Subj:   Tactical queries missing SLA
 
From:   Anomy Anom

<-- Anonymously Posted: Friday, February 26, 2016 14:16 -->

I am having trouble meeting the SLA of tactical queries. We are using applicance 2750, 2 Node system with 120 AMPs, using SLES 11. We have an MQ Series Integrator that sends stored procedure calls to the database. Each Stored Procedure generates approximately 100-120 single amp tactical queries (using primary index) these are mostly deletes and inserts (apart from some regex functions on the input parameter values passed). Typically we receive around 20,000 messages on a daily basis - around 2000 per hour during the business hours. We are not able to process all these requests on daily basis and queue builds up and gets emptied only during saturdays when the call volume is low.

I implemented TIWM, two workloads, Default for all the queries on the system and tactical for MQSI. But i don?t any improvement at all even after implementing the tactical workload. we are only able to process around 500-800 messages on average per hours (against 2000 received). The average queries (tactical) submitted to database per hour are around 160K transactions. ?Most of these run under a second (typically 0.02 to 0.9 seconds).? I checked resusagespma table and we almost have free cpu available all say (no flow control situation).

Some of things that i can?t get around my head is that, same query for instance:

     Delete from A where PI='xyz'

1) The same query takes some times 0.01 seconds and during some other time 0.8 seconds. Thats a plenty of free cpu time ! (atleast 30% cpu free)

2) The numbers of messages processed per hour keeps changing , during night when there is less activity it goes up 1200 messages per hour and comes down to 500 during day when system is relatively busy.


But when these are tactical queries why is this affected by other work on the system?

Also, i have noticed that many of these queries are having some initial delay, (FirstStepTime-StartTime) Second(4,6)? -> 0.01 to 0.8 seconds and these do not have any corresponding workload delay. How do we interpret this delay ?

Is there anything else i can do to process 2000 messages per hour ?


Thanks in advance.



     
  <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: 24 Jul 2020