If you want the summary output as a data set, you will also want to look into the Proc MEANS option STACKODSOUTPUT.
You are summarizing data in different data sets (tbl=Y_TTT*|L_TTT*) you will need, at a minimum, one use of Proc MEANS per table.
If the variables being summarized in the separate table are exclusive to a table, you probably want to stick to one MEANS per table. However, when producing output data for a wide audience, you would be well served to have the source data set name in the summary results.
If the data set structures are similar enough in 'identity of measure (i.e. customer id and region)' and the measure are in different data sets, perhaps due to policy of splitting data by year or month, you can stack the data sets and use one Proc MEANS on the stack. If you are stacking data sets it is good practice to incorporate the source data set name in the summary results if you need that temporality as a categorical variable. The stacking can be a delivered through a view.
With the proper hardware 2m x 100 summarization will happen quickly.
