At one of my clients, we're in the process of "redesigning" their BW environment. Basically we're doing a technical (only) (re)validation of the current system, meaning we just check whether data flows can be improved purely technically (migrate old flows to 7.x, improve coding in start, end and/or expert routines, rearrange characteristics in dimensions, ...).
In addition, we're creating new process chains and slowly replace the "old" ones (on a functional domain basis) with new ones in the productive "meta" chain.
In order to make sure we're doing things right, we need to check (from time to time) whether our efforts have the required impact. One of the things to check is to see whether the loads are performing the same (or, ideally, better). Hence, we should check the (meta) process chain.
Via ST13 I took the data for our "meta" chain (one chain that contains ALL daily loads) for the p ast 2,5 months.
To have a "better" (read: nicer & cleaner) view, I decided to process this data with Lumira (actually Predictive Analysis since you can't have both on the same pc and I have Predictive Analysis installed on my laptop - but, the visualization part is exactly the same).
Here I filtered out the "failed" loads (the ones in red) as they create too much distortion on the view and also I filtered on the pas t 1.5 months (instead of the entire data set which contains 2.5 months of data).
What do we see immediately? The past weeks our daily load time has increased with approx. 10000 seconds (almost 3 hours). Obviously that is NOT the goal of our project. We see that this has been occurring since September 7th (last Saturday). We did transport some changes on Friday (a bit against the rules, but our exception was approved) impacting one flow (concerning PP data).
I already identified where the "error" is located at, namely: one InfoPackage is extremely bigger than the others (in the past this had been split up into smaller InfoPackages, however they did it directly in the productive environment, so those changes were never made in our development environment).
This is just a very "practical" way of using Lumira (or Predictive Analysis) to help identifying (possible) issues in your BW loads.
In addition, we're creating new process chains and slowly replace the "old" ones (on a functional domain basis) with new ones in the productive "meta" chain.
In order to make sure we're doing things right, we need to check (from time to time) whether our efforts have the required impact. One of the things to check is to see whether the loads are performing the same (or, ideally, better). Hence, we should check the (meta) process chain.
Via ST13 I took the data for our "meta" chain (one chain that contains ALL daily loads) for the p ast 2,5 months.
To have a "better" (read: nicer & cleaner) view, I decided to process this data with Lumira (actually Predictive Analysis since you can't have both on the same pc and I have Predictive Analysis installed on my laptop - but, the visualization part is exactly the same).
Here I filtered out the "failed" loads (the ones in red) as they create too much distortion on the view and also I filtered on the pas t 1.5 months (instead of the entire data set which contains 2.5 months of data).
I already identified where the "error" is located at, namely: one InfoPackage is extremely bigger than the others (in the past this had been split up into smaller InfoPackages, however they did it directly in the productive environment, so those changes were never made in our development environment).
No comments:
Post a Comment