How to support BI loading as a basis admin?

[Former Member]

I'm a basis admin for years.

I am requested to support massive BI loading from the basis side.

I am not sure what to do except adding some Oracle datafiles on both the source system and the target system.

What else I can do to help BI loading as a basis admin?

Thanks!

[Former Member · Accepted solution]

Hi,

First of all, there will be a lot of network activity between the systems. When the extraction is first started, the source system will open start processing packages of data with the help of database cursors. As the individual packages are compelte, they are sent to BI. BI stores them in the PSA.

Once all the packages are across, BI reads the PSA and processes them into the data target. Part of that processing will be to generate the surrogate IDs (SIDs) for faster reporting performance. The data is also denormalized. That last part is the most resource intensive on the system.MOst of the process related things are the responsibility of the BI admin. The Basis admin is more related to the work processes and system resources. Ideally, the basis and bi admins work together monitoring their respective processes to make optimal use of system resources with number of work processes and load parallelization so as to complete the loads as quickly as possible.

Basis also needs to make sure the RFC connections between the system remain intact and that the qRFC processes remain registered. Most BI admins don't know much about what happens in the actual RFC communication process, so that usually falls to basis.Also, in times of really huge loads, basis may need to increase the transaction logs / undo table space.

Let me know if you have any more questions

Regards,

Pavan

[Former Member · Accepted solution]

Thanks!

Only 2 more details to ask:

1) add the same amount (or same total size) of datafiles to the source and the target, OR the target needs more datafiles?

2) which t-code to configure load parallelization ?

Any oss notes or how-to guides out there?

Thanks!

[Former Member · Accepted solution]

1) add the same amount (or same total size) of datafiles to the source and the target, OR the target needs more datafiles?

Means the size of the datafiles on the target system need to be more.

2) which t-code to configure load parallelization ?

That is done with process chains via RSPC

Regards,

Pavan

[Former Member · Accepted solution]

Please clarify:

1) you mean the target NEEDS more space than the source?

2) on RSPC, I do not see the local parallazation .

Thanks!

[Diggeshhjoshi · Accepted solution]

Ashley,

Beside what Pavan has said, You will need to watch Tablespace on ODSD / ODSI and FACTD / FACTI. I have observed these tables fill up pretty fast during heavy dataload.

Thanks,

Digesh Joshi

[Former Member · Accepted solution]

1) you mean the target NEEDS more space than the source?

Yes

2) on RSPC, I do not see the local parallazation .

Essentially, the BI admins put the Info Packages and other processes in the process chain and connect them with lines. You can have connections that are "on success", "on error", and "always". By manipulating the connections and using some boolean operators, you can affect when things will run.The only part that is explicitly scheduled is the start process at the top of the chain. Everything else is triggered by the previous process completing.you can even call process chains from process chains.

If you look in SM37, you will see a whole bunch of "BI_*" jobs in scheduled status. Those correspond to the steps in the chain. There will be a "BI_PROCESS_TRIGGER" that is released. That corresponds to the process chain start process.The BI_PROCESS_TRIGGER job simply releases all of it's dependent jobs. As each dependent job finishes, it releases the next job until all jobs are done. Once complete, a new set op jobs are created for the next scheduled chain execution.

Regards,

pavan