You may already be aware of there now being a number of online articles and blog posts concerning issues around the potential for accidental enabling of the new Oracle Database In-Memory Option; with pricing understood to be the same as for Oracle’s Real Application Clusters and Oracle’s legendary tough stance on license audit findings, it is not difficult to see why this has got the attention it has.
The initial InformationWeek Report: Oracle Patch Turns On $23,000 Upgrade, was followed on 28th May by CBR Online reporting Oracle Denies Its £14,000 In-Memory Option Activates By Default and then PCWorld.com Oracle hits back at ex-employee’s claims over in-memory database option.
My years in the IT industry certainly mean it is no surprise to me that my colleague at EMC, Kevin Closson, having raised awareness of the issues on his now 4-part personal blog, nevertheless I am disappointed that he has been subjected to criticism and occasional ridicule from so-called IT industry experts and apparently from some of his former colleagues at Oracle Corporation.
In her blog post Getting started with Oracle Database In-Memory Part I, Maria Colgan, Oracle’s Product Manager for Oracle Database In-Memory, has attempted to address the issues raised without mentioning them directly, by starting with the intention to answer the question “how and when is Database In-Memory installed and enabled”.
My initial response was enthusiastic, I still believe Maria has made a positive contribution to the discussion, I commented as such on her blog post, as I believed she had provided clarity on how to avoid accidentally enabling the new Oracle Database In-Memory Option and as a result avoid exposure to potential license audit problems. However, it has become clear after my own testing that Maria and her colleagues need to look at this again.
This is due to what I have described as “no rows selected” == “DOUBT” – the SQL SELECT statement used to enquire of the feature usage in the Oracle Database makes use of a dba_feature_usage_statistics view that joins 3 WRI tables and in my tests one of these tables (WRI$_DBU_FEATURE_USAGE) is empty!
After much consternation and multiple install and test runs, I concluded that in my initial tests the WRI$_DBU_FEATURE_USAGE table is unpopulated when performed in the CDB$ROOT database and certainly results in the “no rows selected” output (Ed. happy to be corrected here)
However, when the tests are performed using the orclpdb PDB, installed as part of the Typical Installation of the Oracle Database 12c 18.104.22.168 software, I see output similar to that provided by Kevin Closson. I say similar, as you can see below I am yet to reproduce his output that has resulted in such controversy!
Clearly, there is something different about our tests, perhaps some sqlplus flags or other settings. What we need to do is share the exact configuration and commands and get to the bottom of this in a mutually agreeable and respectful manner, let’s see what we can do.