Internals

Next Tuesday (19th September) I am doing a free webinar for ProHuddle. It lasts under an hour and is an introduction to how some of the core parts of the Oracle RDBMS work, I call it “The Heart of Oracle: How the Core RDBMS Works”. Yes, I try and explain all of the core Oracle RDBMS in under an hour! I’m told I just about manage it. You can see details of the event and register for it here. I’ve done this talk a few times at conferences now and I really like doing it, partly as it seems to go down so well and people give me good feedback about it (and occasionally bad feedback, but I’ll get on to that).

This post is about the decision the Oracle database engine makes when it is using a full segment scan approach. The choices the engine has is to store the blocks that are physically read in the buffercache, or read the blocks into the process’ PGA. The first choice is what I refer to as a ‘buffered read’, which places the block in the database buffercache so the process itself and other processes can bypass the physical read and use the block from the cache, until the block is evicted from the cache. The second choice is what is commonly referred to as ‘direct path read’, which places the blocks physically read into the process’ PGA, which means the read blocks are stored for only a short duration and is not shared with other processes.

This is the second part of a blogpost about Postgresql database block internals. If you found this blogpost, and are interested in getting started with it, please read the first part, and then continue with this post.
I am doing the investigations on Oracle Linux 7u3 with postgres 9.6 (both the latest versions when this blogpost was written).

In the first part I talked about the pageinspect extension, and investigated the page header and line pointer array. This blogpost looks at the actual tuples, including the index, and how these are stored in the pages.

This blogpost is the result of me looking into how postgres works, and specifically the database blocks. The inspiration and essence of this blogpost comes from two blogs from Jeremiah Peschka: https://facility9.com/2011/03/postgresql-row-storage-fundamentals/ and https://facility9.com/2011/04/postgresql-update-internals/
I am using Oracle Linux 7u3 and postgres 9.6 (current versions when this blogpost was written).

Postgres is already installed, and a database cluster is already running. Let’s create a database ‘test’ for the sake of our tests:

$ createdb test

Once the database is created, logging on is done with ‘psql’:

I described in an earlier post on AWR views how the dictionary views were using metadata and object links to show information from other containers. But this mechanism cannot work for fixed views (aka V$) because they don’t have their definition in the dictionary.

The big difference is that most of V$ views are available long before the dictionary is opened or even created. Just start an instance in NOMOUNT and you can query the V$ views. Even in multitenant, you can switch to different containers in MOUNT, and query V$ views, when no dictionary is opened.

On 31st of May in Düsseldorf, at DOAG Datenbank, I’ll talk about transportable tablespaces and pluggable databases. Both methods are transporting data physically, the difference is in the transport of the metadata, which can be more flexible when transported logically, as with TTS, but faster when transported physically with PDB. I have a lot of demos to show transportable tablespaces with RMAN, and the different cloning features available in 12cR2. If I have time I’ll show what is inside the dumpfile when using Data Pump to export the metadata. Here is the idea.

expdp transport_tablespaces

Here is how we export metadata with Data Pump for transportable tablespaces.

Actually, this is a follow up post from my performance deep dive into tablespace encryption. After having investigated how tablespace encryption works, this blogpost is looking at the other encryption option, column encryption. A conclusion that can be shared upfront is that despite they basically perform the same function, the implementation and performance consequences are quite different.

Thanks to suggestions made by Frits Hoogland, I made some improvements to the ODBV. The new version can be found here: http://ora-600.pl/oinstall/odbv.x86_64

The changes are:

• Recognition of first, second and third level bitmap block
• Recognition of pagetable segment header
• Block number ranges on the left side

The blocks will be coloured properly to belonging segment.

Thanks to suggestions made by Frits Hoogland, I made some improvements to the ODBV. The new version can be found here: http://ora-600.pl/oinstall/odbv.x86_64

The changes are:

• Recognition of first, second and third level bitmap block
• Recognition of pagetable segment header
• Block number ranges on the left side

The blocks will be coloured properly to belonging segment.

When sifting through a sql_trace file from Oracle version 12.2, I noticed a new wait event: ‘PGA memory operation’:

WAIT #0x7ff225353470: nam='PGA memory operation' ela= 16 p1=131072 p2=0 p3=0 obj#=484 tim=15648003957

The current documentation has no description for it. Let’s see what V$EVENT_NAME says:

SQL> select event#, name, parameter1, parameter2, parameter3, wait_class 
  2  from v$event_name where name = 'PGA memory operation';

EVENT# NAME                                  PARAMETER1 PARAMETER2 PARAMETER3 WAIT_CLASS
------ ------------------------------------- ---------- ---------- ---------- ---------------
   524 PGA memory operation                                                   Other

Well, that doesn’t help…