i am facing issue with TASM behaviour. i realized that tactical workload, with rp of 50%, give results in 2-3 secs, while same workload at busy time give result in 30-40 secs.
i need to know if system is running 100% and tactical query hit system then how much time is reuiqred for tactical to take some cpu back.
It's a very internal question I guess and depends a lot on TASM internals and many factors: your TASM rules, filters, limits/thresholds for the load running on your system.
Other than that are you asking for a solution for improvement here? or are you asking how to monitor this scenario?
TASM will check each job at the end of each step, or at the delaytime setting set in the TASM manager. i need to know how to set these settings, i want TASM to check settings every 10 secs.
Sorry Umer but I could not understand your query clearly. Could you please elaborate what you are referring to as a job? and what do you mean by 'I want TASM to check settings every 10 secs.'? which settings?
i have a user assigned to tactical workload, and tactical rp is 50%, only 1 AG is assigned to this RP.
this user got results in 2-3 secs when system in under utilized.
but when system is 100% busy and then this user starts suffering, and starts getting results in 30secs.
so i beleive there is some aging settings which check system utilization every x seconds and then assigned wait to active workloads.
As far as I know there are no ageing settings for your purpose. There are settings for logging, event intervals though but they don't effect your situation.
Could it be that the SysCon/OpEnv changes when your system is highly utilized and the 50% allocation to the tactical RP is changed?
If thats not the case then it means that your user tactical user's CPU utilization > 50% because if it was below or equal then TDWM should allocate it just enough CPU to execute things immediately but if it goes above 50% at one point in time then the queries/load will consume more time executing.
It's better that you check the utilization/consumption pattern of your users falling in the tactical workload. You can observe DBQL data for that purpose.
Moreover there are tools that can help you identify such situations in advance e.g. BEZVision (http://www.bez.com), Viewpoint
After you have observed your system's behavior for the user in high CPU conditions you can then alter your TDWM settings e.g. give different weights in the priority scheduler, apply a session/query level threshold for the user etc.
First thing that comes to mind reading over this is that either the tactical workload is not truly tactical in nature:
(1) 1 or 2 AMP operation using PI or USI.
(2) Some AMP operation joining on common PI and leveraging NUSI/USI
(3) Well formed ALL AMP querying joining on common PI leveraging highly selective NUSI
or the system is saturated and TASM configuration doesnt allow you to satisfy new work effectively.
Either way the solution may not be simple without fully understanding the nature of your "tactical" work, your TASM configuration, and a few other variables about your environment. Monisiqbal's recommendations are good starting points in the absence of understanding the complete picture.
please check the TASM settings using schmon -d and then use schmon -M to monitor the resource usage across all the AGs and RPs. That will help you detect how much you are actually using and when
Check if you are exhausting AWT's. You might not get the desired response time if there is a a skewed workload on the AMP where tactical work is expecting data. As always, Teradata is as fast as its slowest performing AMP. V12 has pretty nice UDF, that you could use to check AWT. But you could get only count for the highest 5 Amps. You may use puma/awtmon to get more relevant information.
The other piece that you may want to check is the availability of CPU cycles for you to complete the task.