Teradata Monitoring

Database
Enthusiast

Teradata Monitoring

Hi Folks
We are Oracle folks undergoing corporate Teradata Training ( but it was not comprehensive enough ) and unlike Oracle , Teradata Monitoring is really a bit complex. I was monitoring the System using TM ( TTU V12) .
I am still not able to make sense of many things in there and how to interpret these in terms of
signs.Can someone PL help ?

- There are some arcmain session I see that say AMP state as Active and PE as Idle. What does this mean ?

- Under Nodes : Disk Q Length - it keeps changing rapidly from -9000 some to 0 . what does this mean.

- What is the difference between host and disk reads

- AMP worker task and significance

-message Q length

The TM user manual gives 1 liners for these but those dont make sense in terms of interpretation of stuff.

Thanks in advance for your help
Sam
1 REPLY
Enthusiast

Re: Teradata Monitoring

Hi Sam,

- There are some arcmain session I see that say AMP state as Active and PE as Idle. What does this mean ?

[font=Courier New]Amp state active- Amp is nothing but access module processor and it has functions like used for distribution of data,retriving from disks, locking,aggregation,formating etc. In oracle we can compare this to 5 mandate background processes. Since terdata says it has a Shared nothing architecture it means each amp is responsible for its own V-disks ie each amp hold some portion of the table.if the amp is not active ie offline that means it cannot access its subset of table data( again here in Teradata has no single point of failure).u can aslo make the amp as no fallback which is differnt from the default and hope in training u would have understood the concepts of Fallbak and no fallback..

PE has wide range of functions Session controlling , parsing ,optimising ,generationg , aborting etc..
A session is established when a user connects it to the database. suppose a user fires a query and immediately u check the session stae it may see as active, dispatching,parsing or aborting etc. when no query is fired but still is connected the PE has nothing to perform but remain idle.Thats why u may see certain sesssions PE is in Idle state.[/font]

- Under Nodes : Disk Q Length - it keeps changing rapidly from -9000 some to 0 . what does this mean.
---> will have to check [color=*][/color]

- What is the difference between host and disk reads
---> In teradata you must have come across FSG cache, In oracle its like DB_buffer _cache. So wat ever the blocks are being read and not found in FSG cache is nothing but a disk read where as when repeated queries find the data in memory its a Host read.
in Teradata normally production env runs on MPP that is combination of SMP(single module processor).a SMP is called as node and node compromises of Hardware and software components. each node has its own OS and is assigned some GB of memory in parts to OS,amps,PDE and remain memory is used as FSG cache.

- AMP worker task and significance
very very Imp in Teradata. AWT's(amp worker tasks) are the actual workforce which does the work .u might have heard amp retrive data , all amp scan etc. in each amp there are 80 AWT's which takes up the request and does the work .once the work is done it is avilable. back to the amp. for example .. u come out of the Airport and u find a Taxi waiting to serve u. once the taxi leaves to your place it again comes back to the airport to carry other visitors..now visitors are the query and taxi's are the Awt . when thers no Taxi u got to wait and same thing happebs in Teradata. when AWT's are servicing other requests your query which might return the answer set in 2secs may bound to wait intil it finds an AWT to serve it.

-message Q length

--- > will chck it out

The TM user manual gives 1 liners for these but those dont make sense in terms of interpretation of stuff.

Thanks in advance for your help
Sam