Head of Business Analytics & Technology
Why SAP HANA 1.0 SP03 - Project Orange - will be a runaway success
01 Nov 2011
HANA, In-Memory, Business Intelligence (BI), Consumer Business
On November 7th, SAP will release the new version of its in-memory database - the much heralded Project Orange, or SAP HANA 1.0 SP03. They'll probably change the name before I publish this article, but hey.
Project Orange is huge - because it's not just the kind of theoretical analytics appliance that we have available with the current SAP HANA 1.0 SP02. It's a complete replacement for the Microsoft, Oracle or IBM database that SAP requires to underpin its Business Information Warehouse, SAP NetWeaver BW.
Around this, a vigorous discussion took place on Twitter between Mike Bestvina (@techdisruptive), Jon Reed (@jonerp) and myself. It surrounded pricing: @techdistruptive: @applebyj let's say I have a 5TB db and already have BWA. So £2 million for ETL to be sped up??
Well this paints an interesting question. The answer is that on the face of it, it sounds like a terrible value proposition. So where is the value in HANA 1.0 SP03?
What's the price anyhow?
Well SAP HANA 1.0 SP02 is priced by the 64GB increment and pricing isn't publicly available - but it is very expensive. Given that BW databases are typically about 5-10x the size of the useful data in them (this is normal in a datawarehouse as there is duplication due to the way data is organized), customers aren't going to want to pay for 80x 64GB HANA units at the current price.
I'm sure SAP know this and I very much doubt they will price HANA for BW at the same price as the full-fat appliance. But pricing hasn't been released and won't be until the product is available. We will see. If SAP have sense, they will look to do market penetration rather than high-value.
What are the benefits of running SAP HANA 1.0 SP03 for BW?
For my money, the benefits are huge. The headlines are here:
- Massively improved dataload time (a key business pain point for organizations trying to get data loaded before 8am).
- Improvement in query performance even compared to BWA 7.2 because it collapses the database, application server and OLAP engine into a single item. Most customers don't have BWA and if they do they aren't running the 7.2 version which provides benefits.
- Massive improvement in Integrated Planning performance because the planning engine is written in C++ in-memory. Expect 3-5x faster planning functions and even bigger improvement in planning query performance.
- Improved organization agility due to being able to move faster, develop faster and collapse some data layers and replace them with logical layers.No requirement for separate application and database servers. Simplified hardware structure.
- No need for very expensive enterprise disk storage or expensive UNIX boxes. I've seen BW systems that cost millions of dollars. HANA will be far cheaper than this for the bigger customers. It may be a bit more expensive for the SME market, but HANA is not yet a SME value proposition
So what's stopping us going there?
The big question right now is product maturity. HANA 1.0 SP03 hasn't been released and most customers value their data warehouse highly. I think customers will want to see the benefits and the stability of the HANA platform first hand.
HANA 1.0 SP03 will at least bring (expected, not confirmed as it's not been released!) high-availability, point-in-time backup and restore, the ability to scale up and out.
So for me, HANA 1.0 SP03 as a replacement to Oracle, IBM or Microsoft databases is a complete no-brainer. The question for customers is when, not if. The factors they will need to take into account when deciding this is:
- How important is my data warehouse stability to me compared to improved performance?
- How am I depreciating my current platform? Can I reuse it elsewhere? Is it shared capacity e.g. IBM pSeries where releasing space may provide value.
- How big are the business pain points around performance?
If you can answer this and the answers suggest that HANA is interesting - go ahead and buy a Test & Demo license and do a trial migration. It's the same heterogeneous platform migration as any other hardware migration, though you need special HANA-certified consultants to do the migration. Seeing may be believing.
John Appleby 15 Dec 2011
Thanks Rama for the vote of confidence. I don't think that outright stability will be a decision making factor, but that maturity will be.
Some customers use BW for some very mission critical stuff, and it is operationally important 24/7. I think those customers, especially those that have very large databases 20-30TB might wait a bit longer before they move to HANA.
Most of the 17,000 implementations though are <5TB, not operationally mission critical e.g. used for month end etc. and therefore are excellent candidates even today.
In 6 months time I'm confident that HANA will be good for 90%+ of the install base.
John Appleby 15 Dec 2011
Hi Rajeev, yes but on the other hand SAP have been saying to split ABAP/JAVA for 6 years so there aren't so many dual stack systems out there.
I'm sure HANA DB as the underlying layer for JAVA will be released soon, but in the meantime you still have to use your usual RDBMS for JAVA systems.
John Appleby 15 Dec 2011
Ameya: that's exactly what I am saying. HANA DB replaces the MSSQL, DB2 or Oracle RDBMS. I am sure they are not idle - see Exalytics as a half-baked example or why Oracle are not.
I'm not sure what the big 3 can do though, because SAP HANA is such a compelling value proposition for BW customers. Even running Oracle in-memory cached is slower than SAP HANA by some way, from what I hear. I don't see how they can compete, in all honesty.
Rama Shankar - Atlanta, GA 15 Dec 2011
AWESOME JOB on the HANA write-ups. I have been working on couple of feasibility efforts to implement HANA with my customers ( huge companies and few mid-size companies) to implement SAP HANA to replace BWA and/or to substitute BWA and the main concern to my customers are stability and cost. After reading your write-ups (especially: “Why SAP HANA 1.0 SP03 - Project Orange - will be a runaway success”, it appears that stability will be a non-issue based on your experience.
I am confident now that I can convince my client to implement HANA when SP03 comes out as GA (general availability).
Thanks for taking all the efforts to share your knowledge on HANA.
Rama Shankar ( SAP BI Solutions Consultanta – Atlanta, GA – USA)
Rajiv Bahl 16 Nov 2011
Great blog John.
Hi All, Here is a high level summary:
With SAP NetWeaver 7.3 SPS05 and SAP HANA SPS3, the SAP HANA database is available for SAP
NetWeaver Business Warehouse (BW).
There are the following implementation scenarios:
• Initial installation of SAP NetWeaver BW 7.3 SPS05 on the SAP HANA database
• Migration of an existing SAP NetWeaver BW installation to the SAP HANA database
Java is currently not supported with the SAP HANA database. If your existing SAP NetWeaver BW
system is a dual-stack system (ABAP+Java), you need to separate the ABAP stack from the Java
stack before the upgrade.
SAP strongly recommends you to follow a parallel approach in which you keep your production landscape
in place for the time being as a fallback. You can implement SAP NetWeaver BW scenario by scenario,
with the assurance that the existing production landscape is still available as a fallback if required. This
approach greatly mitigates risk while simultaneously enabling you to familiarize yourself with the
administration and capabilities of SAP HANA. SAP strongly recommends that you consider the high
availability and backup/recovery procedures of SAP HANA before starting to use it in production systems.
Hope this helps.
Ameya 09 Nov 2011
I am pretty glued to one of the points that you have mentioned.
"It's a complete replacement for the Microsoft, Oracle or IBM database that SAP requires to underpin its Business Information Warehouse, SAP NetWeaver BW"
By this, do you mean that if I am installign a new BI system, I wont be using any of the existing DBs and HANA DB can be used as a total replacement? If yes, then how do you think Microsoft and other organizations will see this? This is clear killing of their products and I am sure they won't stay idle. As you said as of now it is released only for ABAP engine. Is there a possibility that sometime later it is might as well released for Netweaver Java Stack?
Anand Gupta 03 Nov 2011
Thanks for the clarifications.
I had some idea of what was in the kernel, but the news you are bringing about SP3 is quite a radical change. If SAP has really been able to do it, one can only say Bravo !!
John Appleby 03 Nov 2011
I'm not sure you understand this right - might be worth clarifying it.
1) SAP HANA 1.0 is an ANSI-compliant database with some stuff wrapped around it (calculation engine etc.). It is like Oracle, MSSQL or DB2 in that respect: it provides support for standard SQL commands (at a high level).
2) This means that as of SAP HANA 1.0 SP03, SAP have made the SQL engine in HANA compliant with the ABAP OpenSQL standard, and therefore you can export your Oracle database to a flat file and import it into HANA Database.
3) This means that you can run SAP HANA as the database for any ABAP system - though it is initially only supported for BW. In the BW scenario the OLAP calculation engine and IP planning also run natively on HANA (not on ABAP) so you get even better performance for those tasks.
So when SAP HANA 1.0 SP03 is released next week you will immediately be able to run BW on HANA - natively. Whether it's mature enough for very large or mission critical BW systems remains to be seen, but SAP have made huge advances in getting this far.
Anand Gupta 03 Nov 2011
John, nice piece but I wonder what parts of enterprise applications database , Oracle and Microsoft do you think Hana is going to replace? I guess you do not hint at replacing transaction databases but maybe the BW Oracle/ MS SQL.
Also, and I must add, I have not seen any hints of Project Orange and base my conclusion on projecting what I have seen in SP2; I would be skeptical if SAP can be ready with Oracle/ SQL replacement so soon. I think BW has some very advanced modelling capabilities and it may take sometime for SAP to get that converted into In-Memory space.
If I were to guess we would probably replicate all necessary data from SAP in real time and run the extractor program within Hana...which can clearly speed up the data-availability time, but I have a feeling BW has lot more data-modelling capabilities that are dependent (OK not very strictly) on underlying SQL standards. Even assuming that it can be done, yet 4-5 months time seems a bit too short to actually see the wishlist delivered. If has been, great news of the customers. Or has SAP found a smarter and quicker way?
It would be great to check out where are SAP with HANA at SP3.
Wish you were free to publish a gist of whats coming...with a spoiler warning offcourse :)