Usage of Advanced DSO in BW 7.4 on HANA

[Former Member]

Hi,

I am working on an upgrade from BW 7.3 on HANA to BW 7.4(SP9) on HANA. I have gone through the LSA++ architecture and want to replace the classical DSO's and Infocubes with Advanced DSO's. However, I am struggling to put up a strong case to back this as there are a few questions which I am unable to answer, some of which I have listed below.

1. What is the difference between ADSO and OpenODS? ADSO and openODS both support field based reporting and Infobject level reporting .What are the benefits of using and ADSO over an openODS for reporting?
2. Can ADSO act as a  virtual provider i.e can it report directly from an ECC source?
3. What are the partitioning options available in an ADSO?
4. Is there a possibility to write partitioned data to  disks in ADSO to reduce memory?
5. Does ADSO support NLS? I have read an SAP document published in March, 2014 that it  will be made available in future upgrades, but considering that the client is upgrading to BW 7.4 (SP9) I want to know if this is a possibility now.
6. What are the advantages of using and ADSO over a HANA optimized Infocube( which they already have in BW 7.3 on HANA)?
7. What are the advantages of using an ADSO over a HANA optimized DSO( which they already have in BW 7.3 on HANA)?

I would appreciate any constructive responses for the above questions.

Thanks in advance.

Divahar

[Former Member · Accepted solution]

Hi Diva,

1. What is the difference between ADSO and OpenODS? ADSO and openODS both support field based reporting and Infobject level reporting .What are the benefits of using and ADSO over an openODS for reporting?

Open ODS view is a virtual provider and ADSO is a info provider.

2. Can ADSO act as a  virtual provider i.e can it report directly from an ECC source?

No. By using SLT you can report on a live data but can't be used as a virtual provider.

3. What are the partitioning options available in an ADSO?

check the option "Dynamic Range partitioning" feature for Advanced Data Store Objects. SPO is planned for future release.

As per SAP note   2044468  "Partitioning is only available for tables located in the column store. The row store doesn't support partitioning."  Also recommends partitioning, if there is a risk to reach the 2 billion record limit.

4. Is there a possibility to write partitioned data to disks in ADSO to reduce memory?

A given database table is either memory-based (HANA column or row tables), or disk-based (extended table of dynamic tiering). It is not possible to create a single table that has partitions in memory and partitions in dynamic tiering.  Extended tables cannot be partitioned. When a partitioned in-memory column table is converted to an extended table (ALTER TABLE … USING EXTENDED STORAGE), the partitioning specification is dropped and the extended table is created as a non-partitioned table that contains the contents of all partitions of the original table.

5. As far as I know the request handling technique is Different b/w aDSO and In-Memory Cube or In-Memory DSO. Also in-bound queue in aDSO can be kept in the WARM data. 

Regards,

Shakthi Raj Natarajan.

[Former Member · Accepted solution]

Hi Shakthi,

Thanks for your reply. I have a few questions on your reply

1. My question was more intended to know what are the benefits of an OpenODS view over a ADSO. You say OpenODS view is a virtual provider which means I can just chose to have a virtual provider which will save memory and gives real time information. Please can you explain why I should chose an ADSO over an OpenODS view.

2. When you say ADSO can report on live data but cannot act as a Virtual provider, does that mean ADSO will store the data physically?

Divahar

[Former Member · Accepted solution]

Hello Diva,

If you are extracting the data from ECC to BW in first level we will use the Open ODS. As per your project requirement you don't required any data manipulation we can use this Open ODS directly in reporting. any data manipulation required we will created the ADSO on top of Open ODS. Why we will go for ADSO on top of Open ods.if we will use the ADSO the report performance will be good.

ADSO is not virtual provider. it's an info provider it holds the data physically.

Regards

Sudhir Saripalli

[Former Member · Accepted solution]

Hi Shakthi,

thanks for your answer, it was of great help! However, I'm still confused about the partitioning of tables.

In your answer on question 3 you state the link to SAP note 2044468 that contains the scenario "Tables with hot and cold data". It says there that tables containing hot and cold data can be split.

In your answer on question 4 you state that it is not possible to create a partitioned (extended) table containing hot and warm data.

My question now is whether it is as I've understood that it is possible to partition into hot and cold but it is not possible to partition into hot and warm data (using extended tables) or not. It would be very helpful if you could clarify. Thanks!

Regards

Josef Kaltenböck

[Former Member · Accepted solution]

Hi Josef,

Question 3 was about physical partitioning.(like info cube partition or DB2 MDC clustering)

Question 4 was about splitting data partially in-memory and remaining data in disk.

In general if we talk about multi-temperature memory strategy or dynamic tiering, we need to understand difference between Hot, warm and cold.

HOT data:
Data that is accessed very often used for reporting or for processes in Data Warehouse Management. (queries for InfoCubes, DataStore objects, Advanced DSO)

WARM data: ( Non active Data or Extended tables or EARLY UNLOAD)
This data is rarely accessed. (write-optimized DataStore or Persistent Staging Areas)

COLD data: ( Near line storage or online archiving )
Data of a BW system that is no longer required and maintained for legal purpose, thent Nearline Storage can be used.

If the HANA server has sufficient main memory, then non-active data is available directly in the main memory. When there are bottlenecks in the main memory, it is preferably warm data is removed from the main memory and stored in the HANA disk.

In dynamic process it moves the content from memory to disk for the selected objects. 

Cold Data deletes the data from HOT ( in-memory ) and writes to separate Sybase database. Once data is archived then reports can access the information without loaded the data back into main memory. 

Thanks,

Shakthi Raj Natarajan

[Former Member]

- I want to comment on  direction on ADSO in the upcoming release this december(Planned Innovations). All roads seem to lean towards ADSO and also conversion of Infocubes as well. See the excellent blog http://scn.sap.com/docs/DOC-60425 by

SAP is providing a conversion tool and hence don't convert all standard DSO's but would recommend creating new one's as ADSO.

Now to your bigger question - Can we store some of the ADSO data in Memory and some in Disk and Yes it is possible.

Which means your Hot Data(always updated and frequently used in reporting) in HANA

& Warm Data(In frequent Updates to data and lower latency is OK).

Short summary - Dynamic Tiering( Slide 19 & 23 provide roadmap updates)  + NW 7.4 SP12(Google sap bw 7.40 roadmaps roland kramer if you cannot access the link here) + Data Life Cycle Manager(DLM) available from HANA Datawarehousing foundation 

ADSO seems the logical choice

[RolandKramer]

Hi,

SAP-NLS will support aDSO with 7.50, see details here -

Best Regards Roland

[Former Member]

Hi Krishna,

Have you tried this scenario of splitting data between in-memory and disk?

Only Inbound table in Advanced DSO can be kept in warm data. Data in active table can’t be split between hot ( in-memory ) and warm ( disk ) by using Dynamic Tiering .

Also Check early upload option.

Thanks,

Shakthi Raj Natarajan

[Former Member]

- We are in the process of testing it out .

Issue is it requires about 190 notes when you upgrade from BW 7.4 SP10 to test it out.

Few more weeks and will post an update to this thread.

Thanks,

Krishna

[Former Member]

Hello Diva,

1.What is the difference between ADSO and OpenODS?

Open ODS is the virtual it doesn't holds any data physically. The data will be there is in the data base Data tables and also it's based Modeling not required any info objects. ADSO is physically holds the data and in ADSO we can use up to 120 key fields the reporting performances will be good.

2.Can ADSO act as a  virtual provider i.e can it report directly from an ECC source?

ADSO is not virtual provider it is a info provider. We have create the data flow like info package and transformation DTP we have to load the data from ECC to BW then we can do reporting.

3.What are the partitioning options available in an ADSO?

For ADSO no partition required we can use up to 120 fields as key fileds.

6.What are the advantages of using and ADSO over a HANA optimized Infocube( which they already have in BW 7.3 on HANA)?

Advantages for ADSO we can up to 120  key fields in ADSO's.

7.What are the advantages of using an ADSO over a HANA optimized DSO( which they already have in BW 7.3 on HANA)?

We can create ADSO with change log and or with out change log.Even with out change log also we can load delta to data mart. Disadvantage with ADSO data mart loads will be slow from ADSO.

Regards

Sudhir Saripalli