Today's guest blog is written by Jason Arneil who works on the infrastructure team in Accenture Enkitec Group (AEG).
We work with clients around the work to install, configure, and optimize Oracle Engineered Systems solutions. I started working with Exadata back in 2011 when I was at another company. The original Expert Oracle Exadata book really helped me learn about the Exadata platform. I have been with AEG since March of 2015, and I'm thrilled to be working with the people who wrote what I consider to be one of the best resources on Oracle Exadata. The new 2nd edition that several of my colleagues wrote offers even more insightful information about Exadata. It's available at apress.com and Amazon.
I had the pleasure, recently, of speaking to Tim Mooney, Oracle's Director of Product Marketing for Oracle Management Cloud. He asked me to share my tips for optimizing Exadata. Here's a recap of what I told him:
- Make sure your network is properly configured for Exadata. This is something you can do before Exadata is installed, and it lays the foundation for a successful implementation.
- Resize the default file system on the Exadata compute node. The Oracle Database and grid infrastructure are installed in this file system. As a default, it is set at 100 GB. While this offers a fairly large amount of storage, it will quickly fill up with all of the Oracle loading and homes information. After the device has been up and running for a while and you've done a few patches, the file system can become very full. The 100 GB may not be enough storage. Resizing this file system is an easy thing to do, but it's often overlooked as a point of optimization.
- Decide whether you should enable the write-back flash cache or not. By default, this feature is not enabled. If you have a write-intensive workload, enabling this feature could provide a very beneficial performance improvement. However, if you don't have an intensive workload, this feature won't offer much benefit. You can use the ‘Enkitec edb360 toolset’ to help you determine if your workload could benefit from enabling this feature. This tool helps provide an in depth understanding of what’s going on with a database and to determine if the switching on Write Back Flash would be useful to mitigate certain wait events that would be visible when utilizing the Enkitec tooling.
- Figure out if hybrid columnar compression (HCC) will improve processing speeds for your data sets. This is something you can determine before Exadata is installed. It takes a bit of forethought and planning to make sure it will benefit your data sets. To help you decide whether you should use HCC, you can run Oracle's Compression Advisor tool on your data sets to understand how much processing speed you'll gain from the differing levels of compression. The results will vary from data set to data set, and not all data sets are suitable for HCC. It's ideal for bulk-loaded data, but it's not intended for OLTP or single block read operations.
- Research individual queries and databases to figure out if queries are offloading. While this particular point of optimization requires a deeper level of examination, it can be helpful to improve Exadata performance. You can run scripts to generate a report to figure out which queries aren't following the default off-load feature. Single-row look-ups using an index won't benefit from this feature, but you may find a long-running query that isn't offloading. If you identify such a query, you can drill down into the query to figure out why it isn't off-loading properly. Since this is a more advanced optimization item, Enkitec has developed a process to quickly figure out which queries aren't off-loading.
- Figure out if your application needs to be tuned to run more efficiently on Exadata. Again, this tip requires a deeper level of understanding about Exadata. However, a skilled solution architect with an extensive knowledge of Exadata can examine an application to figure out if it makes sense to tune the application to be more "Exa-friendly."
Most Exadata optimizations are turned on by default. These tips can help you determine how to tune Exadata or your application to maximize computing power. With that said, I recommended one additional strategic tip to help streamline the management of Exadata—use a single IT team to oversee the Exadata environment instead of following the traditional silo'ed management approach with DBAs, system admins, and the like. This will cut down on hand-offs, crossed-wires, and finger-pointing. A single team will help to keep everyone on the same page. Listen to the podcast here.
Jason has worked as a database and systems administrator for more than 16 years, and has worked exclusively with the Exadata platform since 2011. He currently helps organizations manage their infrastructure. During his career, Jason has led Exadata teams for multiple customer projects, been responsible for 50+ Exadata racks, and was involved with all levels of Exadata work. Jason has performed numerous installations, migrations, upgrades and consolidations over several database versions. He has worked as a solutions architect, DBA team lead and systems administrator. His experience spans a broad range of industries, including finance, retail and online web transaction systems. Jason is an Oracle ACE and has presented many times at the UKOUG and at Oracle OpenWorld. He blogs regularly about Exadata at jarneil.wordpress.com and tweets @jarneil.