What is a database?

Question

This question recently arose: "What's an Oracle database?"
It's a common question for someone new to Oracle to have.
I was going to answer in the thread, but I decided to put this thread here, where others can point to it in the future. It's a common question for someone new to Oracle to have. They hear all these words like redo log, tablespace, data file, control file, instance. How does it all fit together??
Well, here's my attempt at covering the basics. I tried to put these in the most logical order I could come up with. Corrections are welcome. I'll edit this as necessary.
The following are common Oracle terminology:
Datafile: An O/S file or raw device, which is used by Oracle to store data. Datafiles are the physical building blocks of a tablespace, which is a logical entity.
Control files: This can be thought of as having all the bootstrap information. It has pointers to datafiles, redo logs, as well as lots of housekeeping information that's beyond the scope of this document.
Redo logs: Redo logs contain a log of all transactions on the database. Whenever a change is made to the database, it's recorded in the redo log.
Database: The collection of data files, control files and redo logs. Taken together, they make up what Oracle refers to as a database. This is where all the data is stored.
Instance: Collection of processes and memory. An instance provides the means to mount, open, and interact with a database. This what you see running on your system, consuming resources and servicing calls from various database clients.
User: An account defined in the database that allows for a means to connect to an instance with a specific set of rights and privileges.
Schema: Collection of objects, all owned by one user. These can include: tables, indexes, synonyms, procedures, functions, packages, types, etc. Usually, all the database "stuff" related to a particular application will be owned by an application owner schema. When a User creates his very first object, this implicitly defines the user's schema.
Segment: An object, such as a table or index, to which space is allocated. Objects such as procedures, synonyms, functions, etc, only consume space in the data dictionary where they are defined, and are not segments.
Tablespace: A logical pool of storage available for use. When a user creates a segment in his schema, he must specify a tablespace from which space will be allocated. A tablespace is made up of one or more physical datafiles (actual files in the O/S).
Extent: A physical chunk of contiguous space, from a datafile. The size of an extent will always be a multiple of the block size. Extent sizes can be fixed (uniform size) or can be chosen by Oracle (autoallocate). When a segment fills the all the space that's currently allocated to it, it will attempt to allocate more space from the tablespace. The size of the allocation will be an extent.
Data block: The smallest addressable chunk of disk space in the database. The block size is chosen at database creation time, and is fixed. (More recent versions of Oracle provide for tablespaces with varying block sizes, but the default block size is chosen at database creation time, and cannot be changed.) The minimum block size is 2k, and the max varies by platform, but is usually 16k or 32k. The most common block sizes are 8k and 16k.
Hope this helps,
-Mark
Message was edited by:
Mark J. Bobak

Answer 1

Information about schemas is stored in a repository called the datadictionary.
The datadictionary is a collection of tables, etc located in the SYS schema.
Every Create table statement etc under the hood modify tables in the SYS schema.
Table definitions are stored in tab$ and obj$, column definitions in col$, etc.
The end-user can query the datadictionary by using a set of views, like user_tables, user_tab_columns, user_objects etc.
Those views are listed in the DICT view, and are documented in the Oracle reference manual.
When you submit a SELECT statement, Oracle (recursively) issues selects on the tables I mentioned above to check whether your query is correct, and how it should be executed. This is referred to as 'parsing'.
The dictionary also contains information regarding the physical location of all tables in the database.
If you want an overview of your own schema, issue in sql*plus
select object_type, object_name
from user_objects
Apart from that, most what you are asking about is described in the Oracle Concepts Manual.
This is a compulsory read.
Sybrand Bakker
Senior Oracle DBA