Clearly, there is a lot of hype surrounding SAP HANA and Oracle Exalytics. They are both in-memory products, and although Exalytics is somewhat newer, HANA has only been generally available for about a year itself. So, we don’t know what either might eventually become, or what newer releases might later bring, and yet there hasn’t been a long enough track record to take either company’s claims at face value.
Both products have shown the ability to speed up analytical queries. But, for transactional applications, there are some important distinctions that would seem to give the edge to HANA. In their recently released internal benchmarks, SAP has shown some impressive numbers (though unverified) with regard to HANA’s scalability. Per the details released InformationWeek, SAP claims that, “100 billion records of relational data, representing five years’ worth of SAP sales and distribution data, were loaded into HANA, which compressed the original 100 terabytes of disk data down to 3.8 terabytes of RAM. The Hana system was running on a commercially available IBM X5 16-node cluster, which is four times as large as the platform used for SAP’s previous HANA scalability test”. For purposes of comparison, it would be very useful have similar benchmarks from Oracle.
SAP has said that HANA is designed to eventually run and do all processing with just one database (although, this has not been demonstrated so far). But, Oracle’s solution clearly requires multiple databases. “Where SAP says a single HANA database will eventually run both BW and core transactional applications, Oracle Exalytics is an add-on product (for all but the smallest data marts with less than a terabyte of data). That means an SAP customer running Oracle will still need one database license for the transactional database, a separate license for the data warehouse database, and a third database license (either TimesTen, Essbase, or, in some cases, both) for Oracle Exalytics.”
The goal, Oracle says, is to get more out of existing databases, not to replace them.
Oracle does offer the benefit of being able to execute your existing applications at a higher level of performance, without any modifications. Both companies have said that they will be creating new apps, although HANA is further along on this path. In the world of Exalytics, transactional data is copied from the application layer to the DW and yet again to Exalytics for executing queries. So that implies three layers of servers, storage, integration, and administration of all of the above.
“Not only do you end up with three copies of the data, you have to add the latency required to move data from one copy to the other,” says Gartner analyst Don Feinberg. “The real promise of HANA is that when the transaction is done, the data is instantly in the data warehouse and your analytics become real time.”
The contrast is that Oracle seems to move the data to the appropriate database type, and uses their expertise and proven technology for that particular type of processing. Hana brings their processes to where the data already is, and performs whatever types of processing is needed, right there. Both of these approaches will work; they are just done in different ways. SAP believes that, with less movement of the data, it can perform the functions in a more instantaneous, real-time fashion.
I would also like to take the opportunity to clarify some of the misconceptions about HANA. A key component is HANA’s ability to optimize reads or writes, depending on what is most critical for processing. And, there are also business functions built into the database.
It supports unstructured data analysis, and it has both database columnar text processing, and text-analysis functionality.
It supports SQL and MDX, and can support parallel query execution, and in a high volume.
My previous blogs on this topic has led to some lively debate on the competitive differences. David Hull, for example, has provided some commentary that he believes refute many of the claims that I cited in the previous blog on the Oracle point of view, and Puneet Suppal has added some perspective, as well. It is not my intention to pick a winner in this, but to bring some of the latest information to light to provoke thought and discussion. Like any technology, the best way to settle this is really to test the two products side by side, using your own benchmarks and criteria.
This will continue to be an ongoing discussion, and I look forward to more of your feedback.
