Posts by Marc Fielding
Welcome to the third installment of a series describing the Oracle Exadata platform. In part 1 we talked about hardware components, and in part 2 went on to discuss software. We now move on to how these components are packaged and licensed.
As I work on putting together slides, I’d like to remind you all of my Oracle OpenWorld 2010 presentation, a customer case study migrating a complex data warehouse environment to Exadata. I’ll be presenting along with Michael Aldrich of LinkShare and Aiman al-Khammash of Oracle, and focusing on our own experience with Exadata, lessons learned, and advice for others choosing to implement Exadata.
As of this afternoon, version 188.8.131.52.1 of Oracle’s Exadata storage server software, is out in the wild. This is the first publicly available version of the 184.108.40.206 branch, a major release including a full OS image with an update to Oracle Enterprise Linux 5.5. A number of bugs causing cell server crashes and hangs have been fixed, including 9472035, 9870117, and 9722560.
en creating an index, Oracle versions 10g and above automatically compute optimizer statistics. And even before that, 9i had a COMPUTE STATISTICS clause to accomplish the same thing. Not only does it save the time and effort of running DBMS_STATS, but it also saves the disk I/O involved in such an operation, since all rows are available in the course of index creation.
I’m not aware of any TPC results for Exadata V2. However a TPC-C test was run in late 2009 using a Sun F5100 flash array on Sun T5540 servers, setting a record for the benchmark that still stands. The Oracle wiki has a great overview of the various benchmarks and what they actually measure.
We’re in the process of applying Exadata Bundle Patch 5, and ran into an issue I wanted to share.The system didn’t have a XDB user, and likely because of this, the two “alter package” commands invalidated the dbms_metadata objects. The subsequent recompilations all failed. On advice from Oracle support, we ended up backing out the patch and re-applying without running these “alter package” commands, and confirmed that DBMS_METADATA works fine. So if you don’t have XDB installed, skip the “alter package” commands. I hope the README is updated soon.
My article Making the Most of Oracle Exadata in the August 2010 issue of the NoCOUG Journal has come out. It covers Exadata’s feature set, and then dives deeper, discussing how to make the best use of its capabilities. For those of you not subscribers of the print edition, it’s also available electronically.
I’ll be giving a webinar about Exadata implementation, where I’ll be talking about Exadata features and how best to use them. I’ll also be sharing some lessons learned from my own implementation experience. The webinar will be on Wednesday August 11 at high noon eastern time. Note that this is a change from the previous date.
After covering hardware components of Sun Oracle Database Machine in part 1, our grand tour continues with a look at the software side. With the prominent exceptions of the Exadata storage server software and the Oracle database itself, the software stack is based on well-known and widely used open source products.
Although many electrons have been expended discussing Exadata’s features, storage indexes tend to figure last, with a vague mention of row elimination in heavily clustered data. Even Oracle’s Exadata software user guide devotes barely half a page to them. Unlike the better known smart scanning features though, storage indexes have an important advantage: rather than offloading workload to storage cells, they eliminate the need to do the I/O at all. Here are some sample statistics taken from an actual production system: