SAS Data Integration Studio, DataFlux Data Management Studio, SAS/ACCESS, SAS Data Loader for Hadoop and others

Long Calculation / Cube times

Accepted Solution Solved
Reply
Occasional Contributor
Posts: 10
Accepted Solution

Long Calculation / Cube times

Recently calculation / cube processing times have increased 10 fold and I am trying to work towards a solution. I have sought out reported / unreported errors (e.g. circular account assignments , percentages exceeding / not equating to 100, invalid driver rules disabled / corrected / removed, OutputQtyUE not equal to 0, etc. I have also made reference values as short as possible to ensure intersections are as short as they can be. Can anyone suggest other actions that could be taken?


Accepted Solutions
Solution
‎04-28-2016 10:20 AM
Super User
Posts: 5,255

Re: Long Calculation / Cube times

If it's cube creation, there should be a SAS log from the batch/workspace server that you could look into. Great if you cold compare with older logs where creation time where "normal".

You could temporarily add some system options to your cube program to see the amount of resources the job is using:

options fullstimer msglevel=i;

Some basic monitoring on the OS level usually helps (and verify that there are no concurrent jobs competing for resources).

 

Not sure what ABM 7.2 corresponds in terms of SAS version, but from SAS 9.4 the SAS Environment Manager is the place to start looking at (SAS) system performance.

Data never sleeps

View solution in original post


All Replies
Super User
Posts: 5,255

Re: Long Calculation / Cube times

First: do you refer to cube creation/update, or query?

The measures you list sounds like something outside SAS.

A starting point would be to try analyze SAS logs, and explore the logging possibilities there.

 

And as with all kind of trouble shooting: what has changed?

  • Environment (HW+SW+NW)
  • Data structure
  • Data volume
  • No of users
  • Search patterns
  • ...
Data never sleeps
Occasional Contributor
Posts: 10

Re: Long Calculation / Cube times

Thank you for your reply:

 

Answer to your first question: cube creation/update.

RE your question about measures: this is with SASABM 7.2 so I'm not sure what to say

What are the logging possibilities to choose from and Analyse?

 

As far as I know, there has been no recent change in the:

  • Environment (HW+SW+NW) - I am not aware of changes (recent or not) but I will refer this point to the team supporting this
  • Data structure, Data volume, No of users - I am pretty close to all of these areas and all have remained constant
  • Search patterns - not sure what to say on this - can you please explain what this means

 

What I can say is that recently I have seen more error messages that appear to indicate a more rigorous reporting / 'enforcing' of calculation logic / errors like over 100% driver assignment / values. I'm not sure if this the case, but it would appears that bug fix / updates may have been loaded recently that include updates to codes driving calculation processes. Are you aware of any updates that were recently rolled out to 7.2 & above ABM users ..... perhaps in readiness for a migration to V8 which we are in the process of doing? 

Solution
‎04-28-2016 10:20 AM
Super User
Posts: 5,255

Re: Long Calculation / Cube times

If it's cube creation, there should be a SAS log from the batch/workspace server that you could look into. Great if you cold compare with older logs where creation time where "normal".

You could temporarily add some system options to your cube program to see the amount of resources the job is using:

options fullstimer msglevel=i;

Some basic monitoring on the OS level usually helps (and verify that there are no concurrent jobs competing for resources).

 

Not sure what ABM 7.2 corresponds in terms of SAS version, but from SAS 9.4 the SAS Environment Manager is the place to start looking at (SAS) system performance.

Data never sleeps
☑ This topic is SOLVED.

Need further help from the community? Please ask a new question.

Discussion stats
  • 3 replies
  • 271 views
  • 0 likes
  • 2 in conversation