RMOUG, Day 1, +1

RMOUG Day 2 has started, but there was so much great content yesterday, I don’t know if I’ll remember it all unless I write a few notes here on my learning. My first session of the day was Graham Wood’s session on adaptive thresholds for monitoring in 11g. I didn’t know very much about these … Continue reading “RMOUG, Day 1, +1”

RMOUG Day 2 has started, but there was so much great content yesterday, I don’t know if I’ll remember it all unless I write a few notes here on my learning.

My first session of the day was Graham Wood’s session on adaptive thresholds for monitoring in 11g. I didn’t know very much about these new methods for setting alerts, but they are certain to be useful. Some of the methods for adaptive thresholds were available in 10g, but many enhancements were made in 11g. Most importantly, the cyclic nature of workload has periods that can be automatically determined in 11g whereas in 10g, they had to be specified manually. Graham talked briefly about using DB Time and Average Active Sessions as important metrics for tuning, but for alerting, adaptive thresholds make the most sense. Setting a hard limit means that you will likely miss many issues. If your system is normally 10% utilized overnight, but spends all night at 60% utilization, you would like to know about it. However, during the daytime, if the system normally operates at 75% utilization, a hard alert limit at 80% would have miss the 6x utilization increase from the overnight hours. Adaptive thresholds wouldn’t miss the aberration and would alert you of the 60% utilization in the overnight period. That would allow you time to attempt to resolve the problem before daytime hours start.


At 10:30, I listened to Kevin Closson talk about Exadata. As you might expect if you read Kevin’s blog, the session was quite technical and he detailed much of the architecture of the Exadata Storage Cell and how it interoperates with ASM and the RDBMS. Over the course of the 2 days I’ve been at RMOUG, I’ve been lucky enough to enjoy several conversations with Kevin, so I had sort of a preview of his session already, but I still managed to learn a few things. Specifically, one of the most interesting things I learned was about the I/O Resource Manager (IORM). IORM is the storage counterpart to the Database Resource Manager and allows you to meter or limit I/O in an Exadata cell based on an I/O Resource plan that is managed in the database. The Exadata cell is told about the IORM plan at the time the request is made since the RDBMS piggybacks the IORM plan metadata with the query request. IORM plans can have declarations for transactions, databases, or usage categories.

After lunch, I attended Tom Kyte‘s presentation about reorganizing objects. Tom proceeded to directly address some myths in his usual way with lots of examples. My takeaways were that indexes with monotonically increasing keys (sequences or current date) can benefit from periodic rebuilds. Tables rarely benefit from rebuilding, so business needs are often required to determine if the downtime required to rebuild the table outweigh the benefits of the rebuild. In many cases, a relatively simple test can show how much benefit may be obtained, if any.

In the vendor presentation session, I listened to Tom Kyte present Exadata. While much of the information is the same as I’ve heard in other presentations, it’s always interesting to see someone new present the same information. They often think of different examples and offer their own experiences. As expected, Tom incorporated some of his own examples.

The last session slot of the day was easy for me to choose–Riyaj Shamsudeen presented a session about new features in Oracle Database 11g that are specifically related to performance. He reviewed a large number of topics in a relatively short period. I have many notes, but the main topics were: onilne index rebuild enhancements, invisible indexes, virtual columns, LOB performance improvements, CBO extended statistics (dbms_stats.create_extended_statistics), fine-grained dependency checking, SQL result cache, PL/SQL function cache, PL/SQL INLINE pragma, compound triggers, and adaptive cursor sharing. There was a lot to absorb and Riyaj offered many examples of these features in action.

The end of the day was the reception in the exhibit hall which was nice for visiting and meeting new people. After the reception, the Oracle ACEs gathered for dinner at a local restaurant. Not wanting to miss opportunities to network with the great minds at RMOUG, Robyn Sands, Kevin Closson, Mogens Nørgaard, and I adjourned to the Hyatt lobby for a nitecap and some great conversation to finish the evening.

The day’s events left me exhausted and energized to wake up again for more great information. I’m not sure if Thursday can top Wednesday’s events, but I’m trying! I sure hope I can make it back to RMOUG again next year and even if I left last night, I would have gotten my money’s worth and then some!