topic Re: Optimize Tmap in Talend Studio

Optimize Tmap

rwnetwork — Fri, 21 Feb 2020 21:37:34 GMT

Hello,

Looking for suggestions on my bottleneck within my transformation steps. I attached an image of my design.

The arrow points to the bottleneck input that has about 100mill records coming out of the SQL query.

I'm selecting Load Once against this SQL query.

The input text file data has only 168k rows coming in.

I have my JVM arguments set at -Xms8G and -Xmx16G.

Suggestions on how to avoid the GC memory overhead error and optimize this job?

Please and thank you!!

Re: Optimize Tmap

TRF — Sun, 23 Feb 2020 15:21:50 GMT

Did you try to change the paradigme of the tMap lookup model?

If your big table has a primary key (or and indexed column) and you got it in the text file data, you may try to change to "Reload at each row (cache)" model. After having changed the model, click the green plus sign on the lookup table, set the "globalMap Key" with a name of your choice then drag and drop the corresponding field from the main table (text file date).

It should look like this:

If you still have memory issues or the job take too long to run because of the number of queries it will generate, you may try to generate a dynamic IN clause with unique values issued from your text file data (something like "1,12,45,32,60,..."). The values will be stored into a global variable (use a tJavaFlex component to construct this variable).

Then on a separate subjob you will be able to query the big table to get only the required records with a SELECT statement looking like this:

"SELECT blablabla 
   FROM yourBigTable
  WHERE yourKeyField IN (" + ((String)globalMap.get("yourVariable")) + ")"

Re: Optimize Tmap

rwnetwork — Mon, 24 Feb 2020 15:29:42 GMT

Thank you TRF for your response.

I did attempt some mods to the paradigm of the tmap, but have not tried the cache option yet.

Let me take your suggestions and see if I can implement them. Will advise on my progress.

Thanks again!

Michelle

Re: Optimize Tmap

rwnetwork — Mon, 24 Feb 2020 20:52:37 GMT

Hi TRF,

I'm running my process with a modification of the tmap doing a Reload at each row (cache), however it's sitting there processing and I feel like it's going to be the same processing time without this change.

Therefore, I would like to attempt your 2nd suggestion of generating a dynamic IN clause, which I envision working, but I have a few questions please:

1. Where do I generate the 'IN clause' from my text file?

2. And the tjavaflex component will store these 'IN clause' unique variables, correct?

3. Then I would create a subjub using tas400input to query the Select statement from the tjavaflex into the tmap I already have created?

Thank you very much for your help!

Michelle

Re: Optimize Tmap

TRF — Thu, 27 Feb 2020 07:43:11 GMT

Hi Michelle,

Having an "IN" clause could be a solution if you are sure the query length limit for your RDBMS will not be reached.

Place a tUniqRow after your data source component then a tJavaFlex in which you will:

initialize a global variable (let's say InClause) with the starting "(" in the start part
add current value (maybe enclosed by '') and a comma to InClause in the main part
remove the extra comma from InClause and add ")" in the final part

On the next subjob you will be able to reuse this variable in your WHERE clause:

"SELECT blablabla
WHERE blablabla
AND yourColumn IN " + (String)globalMap.get("InClause")

Re: Optimize Tmap

rwnetwork — Thu, 27 Feb 2020 14:03:34 GMT

Thanks TRF very much.

Let me give this a whirl and will advise.

Michelle

Re: Optimize Tmap

rwnetwork — Thu, 27 Feb 2020 14:53:24 GMT

Hi again,

After attempting, I realized I haven't done many of these things yet in Talend, would you be so kind to to help clarify a few things?

Re: Optimize Tmap

TRF — Thu, 27 Feb 2020 15:00:52 GMT

Feel free to reuse this example:

Re: Optimize Tmap

rwnetwork — Thu, 27 Feb 2020 15:12:39 GMT

Oh thank you very much for an example!

Will advise on solution.

Michelle