SAP, Industry Analysts and Data Warehousing experts all agree – the future is In Memory. SAP, Oracle, HP, Microsoft, IBM and others are all investing serious money into In Memory technologies. The principle is simple – using slow magnetic disks to read information for a Data Warehouse is very inefficient. If you can pull the Data Warehouse In Memory, you can process billions of sales records to provide an aggregated sales figure, and do that whilst providing sub-second response times. What's more, columnar database technology allows information to be compressed better, so you might only need 1/10 of the space In Memory that you needed before.
Is HANA the second coming?
If you have your finger on the pulse of SAP's Business Analytics strategy then you will know that they are betting a lot of money on HANA. What's more, they are putting a huge amount of media and social media hype into it. SAP had excellent earnings in Q4 2010 from Sanjay Poonen's Analytics team and I even saw someone from SAP mention that this was due to HANA. Come now, HANA wasn't even available in Ramp-Up (SAP's customer specific approach to releasing new software) until December 2010.
Why did Data Warehouses exist in the first place?
Let's take a trip down memory lane and ask ourselves why Data Warehouses were build in the first place. Clearly, to support decision making by providing business analytics – but that information was available in the transactional systems e.g. ERP. So why do you need a Data Warehouse? For these 4 major reasons:
ERP Performance. Analytics against OLTP systems like ERP environments slows down transactional processing. So, you build a Data Warehouse and push data from ERP to the DW overnight, freeing up the ERP environment for online processing for e.g. Sales, during the day.
Aggregation. If you have 1 billion sales orders but you just want to know overall sales globally in 2010, you have to sum 1 billion sales orders in ERP. A Data Warehouse can aggregate the sales orders overnight which means that you only do the calculation once, for many uses.
Business Transformations. If you have multiple sources of information, you may have different versions of the truth, which need to be combined into one view. In a Data Warehouse, you can reconcile those by transforming and calculating against the data before presenting it back to the Analyst.
Master Data. Again, especially if you have multiple sources, you need to reconcile e.g. Product master data so there is a single view of the truth. The Data Warehouse allows for this – and actually this is a particular strength of SAP NetWeaver BW.
How does HANA approach these 4 major issues?
ERP Performance. For now, HANA doesn't approach this any differently to a traditional Data Warehouse (other than taking a real-time feed in some cases, but that is neither here nor there). That's actually OK, but in future SAP expect to run HANA as the OLTP database, so ERP will run on HANA and you won't need a separate Data Warehouse. I'll talk about this in more detail in another blog, because it deserves an essay of its own.
Aggregation. HANA doesn't aggregate by default because it consumes information in real time. SAP says it is OK because In Memory technology allows HANA to process information fast. SAP claim 450bn (that's 450,000,000,000) records in a matter of seconds and I have certainly seen sub-second performance on over 1bn sales orders in a real world scenario.
Business Transformations. HANA handles business transformations by an abstraction layer where the semantics of business transformations are overlaid to the source data in real time. In addition it is possible to "materialize" a HANA view, by consolidating data back out of HANA and back into memory in a transformed fashion – of course, losing HANA's realtime capability in the process. Ethan Jewett covers this in more detail in this excellent blog 'Musing about semantics in BI'.
Master Data. HANA doesn’t really approach master data and only allows one ERP system to be connected.
As you can see, the problem is that HANA's approach is that it simply doesn't satisfactorily approach all of the reasons that Data Warehouses existed for in the first place. Not yet, with HANA 1.0 at least. We don't have quite enough detail for HANA 1.5 to know for sure
Is HANA a serious replacement for NetWeaver BW?
The simple answer: No – at least not just yet. Its limitations mean that it's only relevant for departmental BI or point solutions – in Large Enterprises at least. I don’t think this is a problem short term because there will still be organizations that see huge benefits from implementing real–time In Memory analytics – and they will use HANA for specific situations alongside their existing strategic Data Warehouse.
But because HANA doesn't yet address the reasons why many organizations implemented a Data Warehouse in the first place, it simply cannot take over as a replacement to BW. And in addition, the market hasn't really understood the Value Proposition for real-time In Memory Analytics – they just talk about fluff like "Business at the Speed of Thought". Until the industry specifics are understood, take-up may be slow.
What's more, SAP has built 10 years of Business Content in BW – complex prebuilt business transformations from the SAP ERP suite. This stuff saves huge amounts of implementation time and it's all gone from HANA. If you want these transformations in HANA, you have to build them yourself, based on a deep knowledge of SAP ERP – at least until SAP rebuild the business transformation logic into HANA.
So to the point: when will HANA be monetizable?
And for that matter, is monetizable a real word? Who knows. But I've variously heard that SAP are expecting $100m of sales and 1000 HANA customers by the end of 2011. SAP is doing a conversion for BWA licenses to HANA licenses which will no doubt attract a lot of customers to upgrade their licenses before they actually move off BWA – and because most SAP customers negotiate a license bundle once a year, it will probably happen at that time.
The key point is that SAP choose to give away NetWeaver BW with their ERP product. HANA brings a new stream of license revenue that will fund future R&D and keep shareholders generally pleased. We shouldn't be afraid of that, but SAP customers just want to increase their operating margin (or whatever) – they don't care about SAP shareholders interests.
But who is the target market for HANA right now, and are there really 1000 sites and $100m of sales revenue to be had in 2011? Not with today's HANA 1.0, but HANA 1.5 promises big things like broad parity with NetWeaver BW. If SAP deliver that early in the year, then customers may buy HANA in their droves – they are starting to get why decision making at the speed of thought is important, as this real-time world comes bearing down. The question is – when will SAP bring a saleable product? Time will tell!