This last week, Bob Evans from SAP wrote a blog entitled The Top 10 Reasons SAP HANA Is Disrupting Larry Ellison’s Grand Plans. I think it’s a great conversation to be had, but in the end (despite having some good content), Bob’s article comes over as Oracle-style willy waving and in my opinion loses credibility.
I was considering writing a response on my corporate blog, but I’m not enough of an expert in Exadata yet to put my company’s name to it. So, here are my musings on the subject.
What I will say is that one thing that Bob has right is that SAP and Oracle have the same long-term goal – to dominate the database market – or perhaps more accurately to reframe the database market into something new. Their approach is similar in only one way though – they are starting with an appliance to meet Data Warehouse needs.
How are Oracle Exadata and SAP HANA similar?
First, they are both are appliances – bundles of hardware and software and services that give you something that will run out the box. This means that the 3 elements: hardware, software and tuning – are highly optimised to work together.
Second, they are both initially optimised for Data Warehousing scenarios. This is partially because they are new technology and SAP and Oracle want to focus on getting it right for those workloads, and partially because Enterprise Data Warehouses often perform badly, so there is a simple business case for fast EDWs.
Third, both scale more or less linearly for large datasets. Performance with large data volumes just isn’t a problem with these technologies. Performance on the other hand is a problem for complex analytics where there is a lot of number crunching and neither Exadata nor HANA have really fixed that yet.
How are Oracle Exadata and SAP HANA fundementally different?
Oracle Exadata is a cell based Oracle appliance. This means that an Exadata Appliance is made up of multiple Oracle Exadata Storage Servers, each of which handles a workload, and part of a response set. So you ask a question of Exadata and the central node chunks stuff up and the Storage Servers crunch out responses. The node compiles those responses into an overall response.
SAP HANA is an in-Memory Appliance. It does break up responses if you have multiple nodes just like Exadata, but that’s not really the point. The point is that even with a single HANA node, you can get blistering performance from a reporting query, because it performs the calculation in main memory.
In addition, HANA has an application layer that sits with the In-Memory Database. This is the killer blow because the first native app is a calculation engine. This means it can do complex calculations without an application layer, which massively improves application performance.
So the key here is that whilst Exadata remains a database, SAP are pushing HANA as an application platform – and with good reason.
What does this mean in terms of what’s available right now?
This is where Oracle are doing well right now. They have over 1000 Exadata clusters out there and expect 3000 by the end of the year and I believe Larry Ellison on that. This is because if you have an Oracle Data Warehouse, you can reasonably expect to move it to Exadata right now, and expect dramatic performance improvements. It’s build on the existing Oracle 11g database and therefore isn’t a dramatic shift – just a bunch of existing components put together with some new plumbing.
SAP on the other hand have a number of Proof of Concept customers and a small number of customers who claim to be already getting some value. Their product just went live last week so that’s to be expected, but equally you can’t just take a SAP Data Warehouse yet and shift it into HANA. You have to do a lot of work (for now) building a new semantic layer for the information you want to produce.
What about the pipeline figures?
It’s worth a moment on those. Larry Ellison told the market that they had a $2bn pipeline and has since stopped talking about numbers. Bill McDermott from SAP has stated an existing $100m pipeline which they expect to increase to $1bn by the end of 2011. Oracle are certainly ahead in the game for now in this respect and the reflects the product maturity which Exadata already has.
What do the big 3 hardware vendors think of all this?
Well I can’t speak for HP, but it must be pretty pissed that Oracle are now building Exadata appliances on the Sun Microsystems hardware that it acquired, rather than the HP hardware that HP spent a load of money coinnovating on.
On the other hand, SAP HANA will run on any certified hardware, and the big 3 – IBM, HP and Dell, have certified hardware already and have built business development teams for SAP HANA. I think you can guess where they are going to be putting their marketing dollars, and it won’t be on Exadata!
To answer Bala Prabahar and George Matthew’s questions around DB2 – I think the Economist’s centenary article on IBM “IBM – 1100100 and counting” nails it – published a few weeks back. Look at the split of hardware, software and services. IBM is used to re-inventing itself and if it becomes more successful in SAP & Oracle Services due to a decline in DB2, it won’t care. Bear in mind we are looking at the macro-level here – HANA will take 5-10 years to change the market.
What about the future?
This is where it gets really interesting for me. What Oracle Exadata does is to substantially improve Oracle Data Warehouse workloads. But, I’m not sure that the parallel approach will work so well for OLTP workloads like the SAP Business Suite or Oracle Financials. But more to the point it’s just a means to make Oracle parallel. I’m not sure that it furthers the market that much, but it does make Oracle a very serious adversory to Teradata.
On the other hand, SAP HANA is only just getting started. HANA 1.0 SP02, which was released last week, is a basic analytics appliance. HANA 1.0 SP03, which will go into beta (SAP call it Ramp-Up) in Q4, will compete more directly with Exadata and allow SAP NetWeaver BW Data Warehouse customers to move their systems in-Memory.
And that’s where it gets interesting, because we get the second HANA native app: planning and what-if analysis. And that’s just the start – SAP are building out analytic apps for HANA for things like Trade Performance Effectiveness and will allow the Buiness Suite and its cloud based ERP offering, byDesign to run on HANA.
At that point, SAP customers won’t need Oracle at all, and can run everythign on HANA, or whatever they call it by then.
Taking this a step forward, there is the elephant in the room that SAP aren’t really talking about yet, because it’s premature. Forget about the tens of thousands of SAP customers that are its existing install base. HANA has a much wider appeal if it is a SQL compliant in-Memory database. It could be used for any workload and any app. And that, Bob, is why Larry Ellison should be scared.