SAP HANA Makes Progress and Threatens Oracle

SAP HANA has garnered a great deal of attention in recent weeks and months. For an overview of the technology and its potential, please check my recent blog on the subject, "The real (potential) impact of SAP HANA."

Since that blog was written, there has been a great deal of news in the SAP HANA world, and the surrounding cosmos as well. Rather than write another long blog on this topic, here is a list of some highlights:

• The pricing of SAP HANA is getting clearer. A low end (128GB to 256GB RAM) SAP HANA appliance could start at around $80K-$100K for the hardware. A few sources inside and outside the company (not under NDA) also indicated to me that the price of SAP HANA software is around $120K at the low end, stretching up to $1M to $2M at the high end. SAP says the pricing is not discountable, but early customers have told me that (at least for them), everything is negotiable. This puts the entry-level SAP HANA pricing at around $250K plus services. A pilot project could be done for less than $300K, with potential for very large business benefits. How hard is it to find a business analytics problem that is worth $300K if it can be completed 10x faster? For many businesses, there is no shortage of such analytics problems, and in some cases SAP HANA will actually perform 100x faster - or even better.

• Pricing for higher end SAP HANA boxes is not as clear. If your data can fit into 1TB of RAM, then these single-system boxes should be OK for you. But how will you know if your data can fit into that memory size? It seems (and I would welcome a knowledgeable person's correction on this if inaccurate) that SAP HANA can't be used when data can't fit entirely into one system's memory, which currently tops out at 2TB (and at much higher pricing - an additional $250K or so, from those few vendors offering that high-end configuration).

• How much data can fit on any SAP HANA configuration? From discussions with a number of current SAP HANA users, this is not always clear, and SAP has some work to do to deliver good sizing tools. On average, users I spoke with report compression ratios of between 4 and 10 times for data - that is, if the data would take up 400GB in an ASCII file, the same data with HANA's columnar compression would take between 40GB and 100GB in SAP HANA (your mileage may vary). Of course, the operating system and other software needs some memory to run, but this size of data should fit fine onto the smallest SAP HANA appliances. In a disk-resident database, you might need dramatically more disk space for such a data warehouse, since space will not be used efficiently (sparse data) and since the indices may take a significant amount of space - even more than the data sometimes!

• How about this for a sizing tool - load all your data on a removable hard drive (ENCRYPTED!!!), and bring it to IBM, HP, Cisco, Fujitsu, Dell, and anyone else offering a SAP HANA appliance. Have them load it up for you on their hardware, and test out your queries. If everything is as simple as some customers and integrators have been telling me (which is pretty close to how simple SAP has been saying it will be), then the hardware partners should be willing to do a pilot like this for free. You'll know exactly how much hardware you'll need, and you'll have a very good idea of its performance!

• As a killer new application, SAP HANA will be certified in November for running SAP BW. All BW reports, extractors, etc., should all run without changes (except MUCH FASTER) on SAP HANA after this release. Let's face it - the overwhelming majority of SAP BW data warehouses are running on Oracle databases. If you have an Oracle database license priced in the range of $800K (and many medium-sized data warehouses are on such licenses), with a 25% annual maintenance fee, you are talking about $200K per year just for the database maintenance. If you could migrate that database over to SAP HANA, you might find that you paid for it in the first year by dropping your Oracle maintenance, and at the same time you may be delivering ten times more users or ten times faster results on reports and analyses. It wouldn't be hard to imagine many Oracle databases being converted over to SAP HANA in the coming year - especially those running under SAP BW (and Business Objects soon too!).

• SAP recently also discussed several analytical applications that will be made available on SAP HANA. These are interesting, but not where the short-term value will be for most customers - this will be SAP BW or Business Objects running on SAP HANA, and ultimately the rest of the SAP Business Suite running on SAP HANA. Or Sybase, for that matter.

• Oracle may be feeling some heat from SAP HANA - Oracle just introduced a low end database appliance offering, with hardware and software bundled for between $100K and $300K. While it is unlikely that this appliance would be able to deliver comparable performance to SAP HANA for those applications most suited to HANA's current capabilities, this appliance would have broader application - for example, being currently suited to running production transactional applications such as SAP.

• At the high end, a SAP HANA appliance might cost on the order of $1M for hardware and $1M for software. An Oracle Exalogic data warehouse solution might cost 10x that number, delivering slower performance, consuming a lot more energy, and taking a lot more space.

• SAP has expanded the mechanisms for loading SAP HANA. You can now use log-based replication (replaying the logs on another database to copy it), and trigger-based replication (having a system notify SAP HANA when data is changed), as well as the previously available ETL-based replication (using BW extractors or other technology to copy data out in bulk/batch). The trigger-based replication mechanism is particularly interesting, because it updates the "data warehouse" (can you even call it that anymore with technology like this?!?) continuously, in near real-time. In addition, the trigger-based replication approach allows for consolidating data from several databases quickly and easily into a single data warehouse. If you have one database for your CRM system and another for your financials, both databases can propagate their updates into a single SAP HANA system for consolidated reporting and analytics. Trigger-based replication supports many-sources-to-many-targets replication.

• Business Objects metadata integration with SAP HANA metadata was not announced at SAP TechEd in Las Vegas, but perhaps will be available soon. Business Objects Explorer can run today on top of SAP HANA, which gives some level of integration that will benefit many SAP customers running SAP Business Objects. The more integration at the metadata level, however, the more appealing SAP HANA will become to Business Objects customers not running SAP systems.
  

Conclusions:

1. SAP HANA is getting to the point where any customer running SAP analytics (especially SAP NetWeaver Business Warehouse aka BW) should be starting or at least planning an evaluation.
2. SAP HANA will make its biggest impact - for customers - initially as an "accelerator" for SAP BW.
3. SAP HANA may pay for itself in reduced database maintenance payments ...
4. ... thus its biggest impact - for vendors - in creating pricing pressure on Oracle RDBMS for data warehouse instances.
5. We may be still a long way from seeing SAP HANA replace Oracle, IBM DB2, or Microsoft SQL Server as the transactional data manager for SAP Business Suite.

For more on the strategic impact of SAP HANA, please see http://www.enterpriseirregulars.com/39209/the-real-potential-impact-of-sap-hana/.

Disclosures:

I worked for SAP for six years, including the period when SAP originally acquired the technology at the heart of HANA. SAP is a client as of the time of this writing.