Oracle Database Administration I key terms

Shared memory region allocated when an instance starts, holding the buffer cache, shared pool, redo log buffer, and other memory structures that all server processes access concurrently.

Private memory area owned by a single server or background process, containing session-specific data such as sort areas, hash join work areas, and cursor state.

SGA component holding the library cache of parsed SQL and PL/SQL plus the data dictionary cache, letting sessions reuse execution plans instead of reparsing statements.

SGA region where copies of data blocks read from datafiles are cached, so repeated access to the same blocks avoids physical I/O.

Part of the shared pool storing parsed SQL statements, PL/SQL code, and their execution plans, keyed by statement text and parsing environment.

Full parsing of a SQL statement that requires syntax and semantic checks against the data dictionary plus generation of a new execution plan, consuming CPU and library cache space.

Reuse of an existing shared cursor when a submitted statement matches cached text and environment, skipping optimization and plan generation.

Process that executes SQL on behalf of a user session, reading blocks into the buffer cache and returning results; dedicated mode gives one process per session, shared mode pools them.

Oracle process such as DBWn, LGWR, CKPT, SMON, PMON, or ARCn that performs instance-wide maintenance tasks independently of any user session.

Background process that writes modified (dirty) buffers from the database buffer cache to datafiles, keeping free buffers available and advancing the checkpoint position.

Background process that writes redo entries from the redo log buffer to the online redo log files, flushing on commit and roughly every three seconds.

Background process that performs instance recovery at startup, coalesces free extents, and cleans up temporary segments left by terminated transactions.

Background process that cleans up after abnormally terminated sessions by rolling back their transactions, releasing locks, and freeing resources.

Event at which all dirty buffers up to a specific system change number are guaranteed written to disk, letting recovery start from that point rather than the whole redo stream.

Monotonically increasing internal clock Oracle stamps on every change, used to order transactions, define read consistency, and bound recovery.

Logical storage container grouping one or more datafiles, used to organize segments and to take objects offline or make them read-only as a unit.

Operating system file belonging to exactly one tablespace that physically stores the data blocks of that tablespace's segments.

Set of extents allocated for a specific schema object such as a table, index, or undo segment; each segment lives in one tablespace.

Contiguous run of data blocks allocated as a single unit inside a datafile when a segment needs more space.

Tablespace whose free-space map is stored in a bitmap inside its own datafiles rather than in the data dictionary, avoiding recursive space management.

Locally managed tablespace option using bitmaps to track block free space, replacing freelists and simplifying tuning of insert performance and concurrency.

Storage parameter reserving a percentage of each block for future row growth, so updates that enlarge rows can stay in the same block instead of migrating.

Movement of an updated row to a new block because the original block lacks free space, leaving a pointer behind and adding I/O for every access.

Storage of a single row across multiple blocks, typically because the row is larger than the free space available in one block.

Tablespace holding undo segments that store before-images of changed data, supporting rollback, read consistency, and flashback features.

Minimum period Oracle tries to keep committed undo data available for consistent reads and flashback queries, subject to undo tablespace space pressure.

Error raised when a long-running query needs undo data that has already been overwritten, so it cannot reconstruct a consistent snapshot of the data.

Guarantee that a query sees data as of a single point in time, reconstructed from undo when blocks change during statement or transaction execution.

Circularly reused set of files recording every change to the database, required for crash recovery and standby synchronization.

Copy of a filled online redo log file saved by the archiver process, forming the basis for media recovery and point-in-time recovery.

Event where LGWR stops writing to the current online redo log group and begins writing to the next, allowing the previous group to be archived.

Statement that makes a transaction's changes permanent and visible, and triggers LGWR to flush the transaction's redo to disk.

Ending a transaction by discarding its changes, using undo records to restore the before-images of modified blocks.

Named set of password and resource limits, such as failed login attempts or idle time, assigned to database users to constrain their sessions.

Named collection of privileges that can be granted to users or other roles, simplifying administration and enabling dynamic privilege changes.

Right to perform a class of actions anywhere in the database, such as CREATE TABLE or SELECT ANY TABLE, typically granted by a DBA.

Right to perform a specific action on a specific schema object, such as SELECT or UPDATE on one named table.

Server parameter files storing instance configuration; the SPFILE is binary and server-maintained, while the text PFILE can be edited manually and used to create an SPFILE.

Startup phase in which Oracle reads the parameter file, allocates the SGA, and starts background processes but does not open control files or datafiles.

Startup phase after NOMOUNT in which the instance opens the control files, learning datafile and redo log locations without opening the database itself.

Final startup phase in which datafiles and online redo logs are opened, making the database available for normal user connections.

Shutdown mode that disconnects sessions, rolls back uncommitted transactions, and writes a checkpoint before closing files, requiring no recovery on next start.

Small binary file recording the database name, datafile and redo log locations, and the current checkpoint, essential for mounting and recovery.

Database mode in which filled online redo logs are archived before reuse, enabling media recovery and online backups.

Restoring lost datafiles from backups and applying archived and online redo to roll them forward to a chosen point in time.

Automatic recovery after an instance crash, using online redo to roll forward committed changes and undo to remove uncommitted ones.

Oracle's built-in backup and recovery utility that manages backup sets, image copies, and restore operations while tracking metadata in the control file or a recovery catalog.

Feature using undo data to display table contents as they appeared at an earlier timestamp or SCN, without restoring a backup.

Separate process that receives incoming connection requests on a network address and hands them to a database instance, spawning server processes as needed.

Short name resolving through naming methods to a connect descriptor containing protocol, host, port, and service name for a database connection.

Resolution method in which client tnsnames.ora files map net service names to connect descriptors, requiring no central directory server.

Built-in repository that periodically captures performance statistics and active session samples, forming the basis for most diagnostic reports.

In-memory and persisted samples of active sessions taken every second, used to identify where time is spent during a specific performance problem.

Advisory tool that analyzes AWR snapshots and reports the highest-impact performance problems plus recommended actions, ranked by estimated benefit.

Scheduled jobs run during maintenance windows, including optimizer statistics gathering, automatic segment advisor, and automatic space advisory.

Metadata describing table and index data distribution, such as row counts and column histograms, used by the optimizer to choose execution plans.

File-based directory structure storing trace files, alert logs, and incident dumps for the database and other Oracle components.

Chronological text record of significant database events such as startup, shutdown, errors, and log switches, stored inside the ADR.