With the advent of HANA, SAP BW is endangered species? – Part -I
I happened to check ‘SAP BI Platform with HANA’ and something struck me SAP BW box was missing, No… rather it was replaced by a new solution SAP HANA. Then next wave of thought came in –
Is SAP BW is endangered species? Considering no of years I have spent working on various versions of SAP BW, It made me uncomfortable. I went on digging further on BW & HANA and found quite a few interesting (not so scary) details about SAP BW and HANA.
*Source of Picture – SAP
As a part of this blog series, I plan to share those with you.
At this point in time would like you leave with the thought – ‘IS SAP BW A ENDANGERED SPECIE’ 🙂 Share your thoughts.
The BW - HANA Relationship
In the meantime, HANA is a complementary data marting technology.
The second view is on comparison between what BW provides as data warehousing platform comparing to IC Studio, which comes with HANA:
1/ Can IC Studio replace BW as a platform to build EDW? It probably can, but still BW is much more rich in functionality for this purpose.
2/ Is BW right analytics platform in all cases? In the past it was the only one from SAP, but now you have a choice to find what fits your needs. I had a customer in the past, who build reporting solution on BW for 20(!) users extracting data from one ERP system, and who needed performance so decided to put BWA just for these 20 folks. At that time it was only technology available from SAP, but today - sure thing - HANA + SBO would fit much better into their needs.
Net, it is not that HANA kills BW. It is that now you have a choice and need to think what you need and what solution fits your needs best and at what cost.
So, please stop these HANA vs BW speculations. Sure, BW needs to evolve into nicer+simpler+better, and I am one of the biggest proponents of that.
Cheers,
-Vitaliy
I must admit. I myself had this very first impression of being jobless once IMDB and HANA came into practical existence (now, it has already come and I still am not jobless 😛 ) I got a first detailed understanding of what were SAP's initiatives in regards to this with many 'Road Ahead' sessions at last year's TechEd (Bangalore). Unfortunately, many questions started running in my head for which, alas, convincing and justified answers were missing. There was no confidence among practitioners, users, customers imbibed which can make them aware and bring them to clarity. As it is rightly said 'an idle mind is a devil's workshop', similarly 'a confused mind is a pessimist's castle'. There is a lot of material where we can find everything about IMDB and HANA. But topics such as how BW and HANA will work in-hand are minimum. If these increases, such speculations will go down and what we will observe is smart 'analytic' implementations incorporating both BW and HANA. Till then, confusion will be there, confidence will be low!!
Regards,
Kunal Gandhi
Second,in a sence, HANA (pricisely In-memory DB) is Database. Roughly saying, there are mainly 3 scenarios in HANA.
1. HANA as the Information base
2. HANA as the database of BW (and ther in-memory applicences.
3. HANA as the database of ERP.
1 is the different type of Information base from BW and sometimes competing to BW. but I think appropriate Data scope of HANA is different from BW. so co-existing solution should be practical for many customers especially who have used BW.
2. is like "super BW"
btw, the definition of DWMS is a bit different academically. DWMS is one type of DBMS and focusing on the Data warehousing functions. example is teradata, essbase, redbrick warehouse... BW is not DBMS and not DWMS, it is , as you say, application.
Based on the premise that solutions are developed based on the limitations or potential of technology available at the time.
1. Will the concepts of OLTP and OLAP disappear?
2. If so will there need to be separate systems to support different infomational needs?
2. If an infocube is really a "group by" solution is there a need for infocubes in the future?
3. Following on from the previous question is there a need for DSO?
4. Once the data is integrated from the source systems maybe Business Objects Explorer is the answer?
The answers have varied from "yes" to "to early to tell" to "I don't know"
It will be interesting
>1. Will the concepts of OLTP and OLAP disappear?
They certainly are converging, but I do not think that we are dealing with black and white situation. If we start analyzing particular cases, there will be strict "transactional" processing and will be pure "analytics" processing, and then cases for everything in between, like "actionable analytics" and "transaction with built-in decision support".
>2. If so will there need to be separate systems to support different informational needs?
Let's first see the world of always-on systems (24x7 without planned downtimes). Once we are in that ideal world, then we can start dreaming about point 1 🙂
>2. If an infocube is really a "group by" solution is there a need for infocubes in the future?
Are we talking about BW infocubes? Then they are star schemas really, and analytics views in IMDB are still star schemas too, although not "extended star schemas" as in BW.
>3. Following on from the previous question is there a need for DSO?
Regular DSO or any DSO (write-optimized, real-time)? DSOs are just tables, and tables are still needed in in-memory. Regular DSO plays important role of delta calculation, and IMDB has the concept of delta already built-in.
>4. Once the data is integrated from the source systems maybe Business Objects Explorer is the answer?
Not sure I got it. If BO Explorer is an answer, than what is the question? 🙂
Cheers,
-Vitaliy
Only from a modeling standpoint. The Analytic views are not stored as star schemas in-memory.
I think one of the reasons that "Dimensional Modeling" has been around since the 60s (Kimball suggests this in the DWT book, first edition) is that it's easy for non-technical people to digest when they lack a tool which frontends this. Most modern BI tools allow a business view to be constructed to make the star schema "even friendlier" but then there are always people in the organization who want to get hands dirty and the star schema makes this possible without having to join tables and such. But that whole discussion quickly "goes religious"
I'm no devotee of "cubes" just to say that they serve a business purpose beyond the direct physical translation you see in RDBMS (or snowflake schemes as you see in BW's approach). We're nearly 40 years into the table as a reporting structure though so maybe the cube thing isn't as important to business. Everyone's mileage seems to vary on this topic to the point of, like I said, "religion".
>>Only from a modeling standpoint. The Analytic views are not stored as star schemas in-memory.
Appreciate your time,
-Vitaliy
@greg_not_so
Jacob Klein, vice president of Data and Analytic Engines for SAP, discusses the results of a benchmark test conducted by SAP and IBM demonstrating the power of SAP In-Memory Appliance software (SAP HANA).
Here is that video: http://wp.me/p1jZZK-5E
He didn't answer all questions about HANA. He doesn't promise a rose garden either.
Thanks,
Bala
Ha ha ha:) LOL. About 5(ok. 6 inches) inches of SQL statement. I see them in OLTP systems all the time.
I stress his thoughts/answers are "more credible" than anything else I've seen(relative) on HANA.
Regards,
Bala
So we forced our system to do no swapping to disk and utilize all this fast mempory, and performed a UNICODE upgrade with 10 TB of data.
Wow it was fast, so when people say in memory computing or HANA I think of course, do it faster!