Killer queries: Track them, find them, fix them

Viewpoint
Teradata Viewpoint is Teradata's strategic and innovative SOV (single operational view) for Teradata DB, Aster, and HDP Hadoop systems management and monitoring that enables Teradata's Unified Data Architecture (UDA).
Teradata Employee

Re: Killer queries: Track them, find them, fix them

Hey Robins.. Thanks !!
Enthusiast

Re: Killer queries: Track them, find them, fix them

Hey Shrity, Good Article ! A question on the CPU % Use. I see many sessions with more than 100% CPU usage. I would expect it to be below 100%. Is my assumption incorrect?
Teradata Employee

Re: Killer queries: Track them, find them, fix them

Hi Shrity,
Can you explain Delta CPU and relate it to CPU Tunning? What is the best way to control high IO utilization? Any tip for tunning IObound system?

Thanks in advance.
Badshah Rehman
Teradata Employee

Re: Killer queries: Track them, find them, fix them

Thanks Vijay. Can you let me know which CPU related metric you are refering to, that will help me in answering the quetion. I assume it is in query monitor portlet.
Teradata Employee

Re: Killer queries: Track them, find them, fix them

Hi Badshah Rehman,

Delta CPU is Total CPU usage time consumed, in seconds, since the last sample. This is a good metric to identify CPU consuming sessions at that point in time, especially if the value is very high it “may” indicate query is executing an expensive product join.

It requires analyzing historical data of system to build a strategy to tune any system. However I think having a good combination of CPU bound queries with IO bound queries can improve resource utilization of a system.
Enthusiast

Re: Killer queries: Track them, find them, fix them

Yes, it's in the Query Monitor Portlet. I'm referring to the CPU Use metric. It's showing numbers that are greater than 100.
Teradata Employee

Re: Killer queries: Track them, find them, fix them

Vijay, it sounds like a bug please open an incident.

Re: Killer queries: Track them, find them, fix them

Great discussion so far. I have few questions...

1. Is there a way to calcaulate Request CPU and AMP CPU Skew from any of the DBQL tables after queries finished running?
2. How to determine (via process or a formulae) if a Teradata system is CPU or I/O bound?
Enthusiast

Re: Killer queries: Track them, find them, fix them

Nice article Shrity
Teradata Employee

Re: Killer queries: Track them, find them, fix them

Ram,

You field guy may have an answer to your questions. I am not an expert in DBQL area so cannot comment.

Thanks,
Shrity