Werner Still

@oow15 - High Availability with new functions

Blog Post created by Werner Still Employee on Oct 27, 2015

It turned out that all sessions that I booked today had a relation to High Availability. With a focus on DB Infrastructure and Data Protection this is not really surprising.


But there is news on the database side. With 12.2 there will be Oracle Elastic Sharding. A very new concept for Oracle to leave the shared disk scenario for a single DB and use shared nothing setups.

Primary target currently are scalability issues. There are lots of databases that need to scale far beyond the traditional Oracle setups. Driven by social media demands NoSQL databases emerged this field. Other databases like SAP HANA used a shared nothing implementation as well. Now Oracle incorporates this way of data handling into the database as well.


New fields for this type of technology will be to use smaller, converged devices (only suited to special applications) and the on-premise / Cloud mixture of databases. Other ideas are to have the data where it belongs, customers of a specific country are hosted in that country, although the overall DB is one big image.


Reducing the technology to the very basic parts (and this is not to embarace the developers here) there are multiple DBs linked via Data Guard or Golden Gate that have a specific Database as the content repository and enhanced listeners as Data Broker. The data is split on the individual DBs via a partitioning like mechanism. There are several ways to define the key for the distribution, it is even possible to have up to 3 layers of distribution keys. To allow the complete query to run purely on a single node some tables will be replicated.


All of that is not for free, at the creation of the tables the type of table and the distribution need to be defined. Adding and removing nodes on the fly is possible, there will be automatic rebalancing activities to handle this.


Where I see a very big impact is the protection of these type of installations. Will it be possible to use "traditional" methods for Data Protection? Do we need totally new ways to protect these data?