on 01-09-2014 2:55 PM
Hi Experts,
On our client BW7-Dev sys- we have to run PC that load data (PSA only) from 4 IP's 2 times/month my questions are :
1- what is the Best (from Performance respective) way to build the IP in PC?(give a look @ attached photo )
2- If we have huge amount of data to be loaded does it affect the sys generally? assuming (1-sys landscape )
3- Do I have to change the date in IP manually before I started the PC every month? (I'm thinking of event PC )
Appreciate your helps
Regards
Mazin
1- what is the Best (from Performance respective) way to build the IP in PC?(give a look @ attached photo )
Parallel is fastest. However, many Dev systems are under powered. I would limit the the number of paths to 3. I would also "stack" the load by "subject" (one path for Material, Material-Plant, another path for Customer, Customer Sales. to limit cross locking. for master data) .
I prefer to load all texts, attributes, and then hierarchies. Loading the texts first helps insure that a SID for the characteristic is available. (Master data uses a lookup then create logic. using text first give th look up something to find.)
2- If we have huge amount of data to be loaded does it affect the sys generally? assuming (1-sys landscape )
Yes, you are consuming processes and i/o (input output or read reads/writes) but not CPU. Get with your basis support to find out how many process you are using. For large loads, consider using one path with parallel processing enabled, running by itself for fastest through put. Make sure the you don't exceed the number of processes available.
3- Do I have to change the date in IP manually before I started the PC every month? (I'm thinking of event PC )
Consider setting up transnational data loads using the delta process. For relatively small loads, it may be faster to delete the data completely from the target InfoProvider at the start of the chain and use a full load tahn to manage a delta.
I hope this helps.
John Hawk
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Dear all Experts M Tibollo ,kodanda pani KV ,Ramanjaneyulu Korrapati ,John Hawk and Asha P
Thanks for your reply its really mean a lot to me ...
For sharing knowledge/experience on this situation, we used 3 instead of 4 IPs (John Hawk) in serial Load (Asha P) because it all came from same source, regardless to our small amount of data.
Again appreciate your helps
Regards
Mazin
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi Mazin,
Scenario 1:- after creating process chain in Development, if the samething has to be moved to production, then below parameters need to be checked;-
1. if it is transaction data laod, and if it has to be triggered on weekday, make sure that there should not be any clash between loads. and try to schedule in such timing, where background processes are bit free.
2. If the process chain has to be scheduled on weekend, and if there are weekend loads running then same above things have to be taken care. If there are no weekend loads then no issue.
3.Coming to performance of the process chain using infopackages, in case, if all the 4 infoapckages are from same source, then go for serial load. in case if each infopackage has different selection parameters then go for parallel load.
4.If each 4 infopackages are from different sources, then go for parellel.
points 3 and 4 can be applied for production and development scenarios.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
1. Your diagram 1 shows as parallel connections.its best option as suggested already.
2. No effect on system. its dev system, so there won't be much loading performance problem compare to prod.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi Mazin,
1- what is the Best (from Performance respective) way to build the IP in PC?
A. if you use multiple info package not an issue, its depends client requirement.
2. If we have huge amount of data to be loaded does it affect the sys generally?
A. it won't effect on system
3.Do I have to change the date in IP manually before I started the PC every month?
A. date changing every month not yet required if process chain will fail give next day date it will work next day onwards.
Thanks,
Phani
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
1. in parallel is the fastest way
2. should not be a problem
3. if selection criteria of the IP change every month use a routine in the infopackage selections
BR,
M.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi M Tibollo,
Thanks for your quick reply. would you explain more about parallel. do I have to determine the type of data update on the Processing tab page or in PC setting ? Sorry I never did this before
Regards
Mazin
Hi Mazin,
In IP processing tab
1.PSA then into data targets(package by package) - means it will load form PSA to Direct target
package by package
2.PSA then into data targets in parallel- (package by package) - it will load parallel if you many request.
3.Only PSA- it will load load source system PSA. ( BI 7.0) - this option is enable.
4.Only Data Targets- it will load data upto target.
Thanks,
Phani.
User | Count |
---|---|
89 | |
10 | |
9 | |
9 | |
9 | |
6 | |
6 | |
5 | |
5 | |
4 |
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.