SAP auth/new_buffering

Question: Does anyone have any experience relating to performance impact of enabling/disabling the user buffer?

As of now, we have a system running with auth/new_buffering set to 4.
I have a feeling that not using the buffer may influence our performance, but it is hard to verify without running traces in the productive system?

If the buffering does indeed affect performance, will this effect be larger or smaller using structural authorizations?

I hope someone can help clarify the issue?

/Morten P. Koehler

Answer:
THe user buffer referenced in AUTH/New_buffering has nothing to do with structural authorizationsa as these are stored in a table ( if turned on) and only retreived if the user encounters a HR authorization check.

THe system impact from performance is based on logon time and the number of records SAP has to retreive from UST04 and UST10 and UST12. If you have a properly designed security it can be minimized but the use of composite and task oriented roles increase the records SAP must resolve at logon. It also impacts the size of the system you must have and the swap space you will need for processing if you have 1000's of records to store in memory. In either case the buffer exists, it has more to do with logon times.

The MAjor imact is opening up your system to a host of back doors id auth/new_buffering is > 0.

SAP's claim is a 4 must be used so you can change the users access without having them log off and back on, does not work 100% of the time.

Also the >0 setting is based on a table that are not in sync with the real security tables. SAP only syncs them once and if you do not sync them yourself ( you have to use the sync function module correctly or nothing happens) you are loading incorrect data and orphaned data.

Answer:
The symptom is "without connection to user". I.e. no logon time.

SAP also mentioned that we could also deactivate the buffering in the data-dictionary, but no client (with John's performance pre-requisities) has been willing to take the step to date. Would that (SE11) have an impact from any experiences?