OOW15 – Day 2- Oracle Database 12cR2 designed for the Cloud

By Franck Pachot

.
Oracle Database 12.2 will come in 2016 with new features. Here are some features that are planned – of course this can change until release.

Pluggable databases

Did you worry about the limit of 252 PDB per CDB? Then you will be happy, the limit will increase to 4096. Of course, this is the kind of new features that are designed for the Cloud. If you have only one single database, then do you care about that? Yes, you should. The non-CDB architecture is deprecated. Of course it’s still supported but it’s not the future. All new features will come to the multi-tenant architecture and we will have to go to it event if it’s only single-tenant.

Lot of improvement here:

• More isolation. Resource manager to limit the memory in addition to CPU and I/O. And in RAC (or RAC One Node) Singleton services associated to a PDB will not have all the interconnect overhead as they don’t need it.
• PDB hot clone/refresh/relocate. Don’t need to put the source in read-only for cloning, can refresh a clone, can relocate a PDB to another CDB online. I suppose that redo and undo management has changed a lot for that. Maybe one thread per PDB.
• Application Container. CDB consolidate the database dictionary with object links and metadata links. If we want to provide same application as multiple PDB, we would like to do the same. That will be possible

In-Memory

Here again, the features that are needed to consider it mature.

• In-Memory column Store on Active Data Guard
• Automated IMCS population. Now the IMCS will be a cache, with aging out and repopulating managed by LRU algorithms. Heat Map that has been introduced in 12.1 will be used in 12.2 for that.

We can expect the population statistics that are used for CDB decision to be exposed without needing a 10053 trace.

And for SPARC M7 the software in silicon: it’s for performance(decompression) and security (isolation of address space)

Other features we can expect

You have seen that the dictionary stores identifiers as VARCHAR(128) since 12c, but they were still limited to 30. That will increase.

Sharding: It’s like partitioning in a shared nothing database. You will have a SDB with tablespace sets, and up to 1000 shards. DDL applied on one shared will be pushed to the others. Some common data can be duplicated.

Want to tune the reporting on Active Data Guard? You will have AWR.

Talking about performance, we can expect an Optimizer Statistics Advisor to apply statistic gathering best practices automatically. Reason is that it’s often badly done and not maintained. That looks like an extension of SQL Plan Directives. It’s free (as in no option – don’t know yet about SE)

Application continuity – the option that makes the application highly available when the database is highly available – will support all clients.

Partitioning improved to support more combination, more online operations on partitions, and easy (and online) partitioning of a non-partitioned table.

Cursor invalidation may be improved. Currently any DDL invalidates all related cursors and this can become more optimal in 12.2

This is information made public at Oracle Open World. It’s probably the last time I publish about 12.2 features before it is released. As with 12.1.0.1 and 12.1.0.2, being beta tester prevents us to communicate about new features before they are officially released.