Infrastructure at your Service

uk1

Today at the UKOUG Super Sunday in Birmingham, I had the opportunity to assist at interesting conferences.

The first presentation was about Oracle RAC internals and its new features in version 12.2.0.1 on Oracle Cloud. The main new features concern the cache fusion, the undo header hash table, the leaf nodes, and the hang manager.

In 12c release 2 in a RAC environment, the cache fusion automatically chooses an optimal path; the cache fusion collects and maintains statistics on the private network, and will use this information to find the optimal path network or disk to serve blocks. We can consider that flash storage will provide better acces time to data than the private network in case of high load.

In order to reduce remote lookups, each instance maintain a hash table of recent transactions (active and commited). So the Undo Header Hash table will improve the scalibility by eliminating remote lookups.

Flex Cluster and leaf nodes were implemented in 12cR1. With 12cR2, it is now possible to run read-only workload on instances running on leaf nodes.

Hang Manager has been introduced with 12cR2. It determines sessions holding resources on which sessions are waiting.  Hang Manager has the possibility to detect hangs across layers.

The second was about the use of strace, perf and gdb. This was a very funny presentation with no slides, only technical demos. It was talking on how to use strace, perf or gdb without being an expert. The speaker showed us the different analysis we can realize with strace gdb or perf in case we realize a sql query over a table in a file system tablespace or an ASM tablespace.

Using those tools allowed us to understand the mechanism of physical read and asynchronous I/O, and showed us the differences between asynchronous I/O and direct path read between ASM and file system.

It showed us that the use of strace and gdb is very simple but not recommended in a production environment.

The last session was talking about dba_feature_usage_statistics, and the speaker describes us the components behind the scene.

This view  as its name indicates it displays information about database feature usage statistics. The view gives an overview of each option pack that have been used in the database and are currently in use. It also provides information when the product was first used and when it was used for the last time.

It is not very easy to find information in the Oracle documentation about how this view is populated. But the speaker gave us important information about wrl$_dbu_usage_sample, wrl$_dbu_feature_usage and wrl$_dbu_feature_metadata which are important for the dba_feature_usage_statistics view.

He also showed us a method to refresh manually the view dba_feature_usage_statistics.

Tomorrow another day of interesting sessions is waiting for us !

 

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Oracle Team
Oracle Team