SAP HANA versus Microsoft Business Intelligence and xVelocity

This will be in a long rant format as I am still trying to organize my thoughts around this subject.

Before I start comparing and contrasting the two vendors, I just wanted to make a quick statement to get something off my chest:

Generally speaking, HANA is a good business decision for SAP as it helps SAP to vertically integrate into its value chain by squeezing out the middle men (Microsoft, Oracle, and IBM) as a platform to run their ERP and/or BI stack. SAP ERP obviously needs a database engine but a lot of people do not know that SAP BW also requires a traditional database engine as it does not have its own multidimensional engine to run queries on. Having said that, from a technology perspective, HANA does not bring anything new to the table – basically everything that HANA does (well, claims to do) can be done with existing, proven and robust technologies (both in terms of hardware and software) often at a fraction of the HANA price tag.

At a high level, HANA is positioned using the following three points:

1. Appliance – SAP partnered up with server hardware vendors (HP, IBM) to manufacture the appliance
2. In-memory – SAP states that commoditization of hardware makes in memory the only solution to meet high performance needs
3. Native ERP integration – SAP claims that HANA will have native connectivity into the ERP to alleviate the data integration hell that every SAP ERP customer has experienced during and post ERP implementation

So let’s look at just these three points for now in a little bit more details and see how HANA compares to Microsoft BI stack.

Appliance: SAP is entering a fairly crowded space of appliance BI where each major vendor (Oracle, IBM and Microsoft) already has a robust appliance solution.  Both hardware and software vendors love appliances for one obvious reason – high margins.  The hardware vendors being the happier of the two given commoditization issues that drive their margins down.  I, however, think that the appliance argument is becoming somewhat oxymoronic as on one hand we observe commoditization and precipitous drops in the cost of hardware while at the same these appliances command multimillion dollar price tags.  Just to have some fun, I spent about half hour on newegg.com pricing out various server configurations, basically, for about $10k I was able to price out a machine with 40 cores and about 1TB of RAM.  Let’s say for about $50k I can build out a complete system with SAN and networking.  Why would anyone pay $1M+ for a similar configuration that ties a company down to a single vendor and a very expensive support agreement is difficult for me to understand.  Microsoft has a similar offering – Parallel Data Warehouse (PDW) that carries a high price tag and is suffering from a similar set of issues.  Luckily, Microsoft Analysis Services runs on commodity hardware and provides sub second response time as a query engine.  I am fairly confident that a $100k commodity server running Analysis Services will comfortably outperform the beefiest multimillion dollar HANA box.

In-memory: this is an interesting one.  I think that a lot of this is driven by the success of Qlikview which is driving where all of the Microsoft BI R&D dollars are going and also, apparently, SAP’s.  Chris Webb wrote enough on the in-memory strategy for Microsoft, so I will not address the in-memory issue per se. The only thing I would like to do is compare HANA with what Microsoft has to offer – xVelocity.  I am a big fan of xVelocity for basically two reasons, a) it works as advertised and b) it’s basically free (with some caveats obviously, one just needs an enterprise license of SQL Server).  Again, I am struggling to understand why would anyone pay millions for an appliance when a similar technology works just as well on a commodity hardware (well actually it works quite a bit better as it allows to swap just in case the amount of data is greater than the amount of RAM available)

Native ERP integration: this is basically a myth.  HANA currently supports trigger and log based data replication, which is ironic because log based replication (for example offered by Simplement) was condemned by SAP until HANA became a reality.  Anybody who has implemented DW and BI initiatives on top of SAP has had to combat the FUD about “are you in business of moving the data around” as SAP labeled moving data outside of BW as a bad practice.  Well, funny how things change, since that is exactly what they are doing now with HANA.  Even though HANA is sold as an in-memory solution, obviously, it needs persistent data storage just in case somebody decides to turn HANA off.  The sizing requirements for the storage are in fact fairly significant because it has to account for all of the typical data warehousing data movement and ETL demands.  SAP also offers accelerators (for example, its COPA accelerator) to facilitate this ETL efforts.  So if I have to do all the ETL and data management work anyway, I am struggling to understand why I would spend millions of dollars on HANA when I can do all of that with the tools and frameworks that I already have in place (SSIS for example).

From the technology perspective, I don’t want to say that HANA does not have any place in the market place.  Today it is not a very stable technology according to my not statistically significant set of references, but it will become more robust with time.  Given a variety of rational but mostly irrational reasons, HANA is likely to become part of the IT landscape.  However, HANA comes in several configurations (at list six at the time of this writing), therefore, I strongly urge the IT decision makers not to over buy.  Most of what HANA brings to the table can be accomplished at the fraction of HANA’s cost using Microsoft BI stack running either in the cloud or on a commodity hardware so using the SAPism, the best practice should be to buy as little as HANA appliance as possible and investing the rest of the savings in building the content for the Business Users which, ultimately, is the only thing that business users really care about.

This turned out to be a lot longer post than I anticipated, but hopefully I have given you some food for thought and good questions materials…