Next-Gen Mission-Critical @ IDF: Oracle Exadata + Intel Xeon

If you come to the Intel Developer Forum this week, be sure to attend the session I will host on the Oracle Exadata platform and the Xeon processor's role in it.  (Session DCPS005, Wednesday 9/15 at 10:00a.m.)

The second generation of the Oracle Exadata platform is now almost a year old.  You might not realize it from looking at the Oracle product descriptions, but from the beginning, Exadata has been constructed from server components built around the Intel Xeon processor.

The first generation of Exadata was built from a collection of two-socket Xeon(r) processor 5500-series server platforms. The central, unique innovation of Exadata (vs. conventional Oracle database deployments) is that it divides the processing of database processing into two elements - the database machine element handles the most complex aspects of data management such as join processing, lock management, etc., while the storage cell element divides the often time-consuming table scanning process across multiple 'storage cell' nodes running in parallel.

Parallel processing is not new to relational databases.  Originally, Teradata came up with the idea in the 80's when they introduced their data warehousing product to accelerate largely mainframe-based database queries (Teradata has used Intel processors from the very beginning).

However, Exadata is the first example of parallel processing applied to Oracle databases, and it is done in a manner that isn't limited to just business intelligence or data warehousing functions.

This is because the database machine component of Exadata is running standard Oracle 11G R2 with Real Application Clusters (RAC), which are available if a database image needs to span multiple processing nodes or high availability is required.  The Smart Scan feature, provided by the parallelized storage cell component of Exadata, comes as an added bonus that enables parallel query acceleration (and eliminates the need for expensive enterprise storage arrays!).

So, what is the role of the Xeon processor in Exadata?

First of all, there's the usual - performance.  Exadata provides a new feature: hybrid columnar compression. That's very useful for query acceleration, but it chews a LOT of processing.  Xeon provides the raw horsepower needed to make HCC practically usable.

Second, as of Exadata V2, Xeon provides scalability in the form of the 8-socket Xeon 7560 processor-based database machine option that was introduced last year.  This year, look for that offering to be upgraded to the next-generation Xeon E7 platform, which offers still more headroom for even larger mission-critical workloads.

In these days of pervasive concern about security, the Xeon E5 processor delivers the encrypt/decrypt performance in the Exadata storage cell element that is required to make the Oracle Transparent Encryption feature practically usable.  Without the AES acceleration capability of the Xeon E5 processor, whole-database encryption is just too costly to be practical.  The Xeon E5 processor makes it deployable.

Those are some of the highlights from the session.  I'll be joined by Sumanta Chatterjee, Oracle VP and a key leader of the of Exadata development team, and Hubert Nueckel, lead optimization engineer from the Intel team that works with Oracle to optimize their system for the Xeon processor.  Hope to see you there!