This is the third post on Database In-Memory (DBIM) columnar format use in Exadata Flash Cache. The first was Columnar Formats in Exadata Flash Cache where I described how Exadata is able to transform Hybrid Columnar Compressed (HCC) data into a pure columnar format starting in 220.127.116.11.0 of the Exadata System Software.
I then went on to describe that in 18.104.22.168.0 of the Exadata System Software HCC data would be transformed into true DBIM format, and that Exadata System Software 18.1.X releases support all types of table data. In addition to my blog post, Roger Macnicol also described how this worked in his series of blog posts, the first of which was titled "What's new in 12.2: CELLMEMORY Part 1".
This was really the start of a very concerted effort on the part of the Database In-Memory development team to leverage Exadata Flash storage as a secondary storage tier for columnar formats to power all types of analytic queries. Exadata Flash storage capacity is measured in Terabytes, not Gigabytes, and in combination with DRAM on the compute servers this provides Exadata customers with much larger analytic capabilities than can be provided by DRAM based column stores alone.
The second blog post was written by my colleague Markus Kissling where he described How to Determine if Columnar Format on Exadata Flash Cache is Being Used. In this post Markus describes how to determine if a query is actually using the columnar format in the Exadata Flash Cache.
I said that the Database In-Memory team has been working to further leverage DBIM formats in the Exadata Flash Cache and this has paid off with the autonomous nature of the Exadata Flash support. Once Database In-Memory has been enabled on Exadata any data that is accessed with a Smart Scan will be automatically formatted into DBIM columnar format in the Exadata Flash Cache starting with the Exadata System Software 18.1.X releases. This means that you do not have to perform an ALTER TABLE … INMEMORY command to take advantage of this capability in Exadata Flash.
Even more exciting is that the Autonomous Data Warehouse (ADW) also uses this columnar format in Exadata Flash and it is automatically enabled under the covers. This is one of the key features that allows workload in ADW to run without all the effort that is typically required to run an on-premises data warehouse.