NCM-DB key terms

The running set of memory structures and background processes that manages a database's files. It exists only while the software is running and disappears when the server shuts down.

The persistent collection of data files, control files, and redo logs stored on disk. It survives instance restarts and is mounted and opened by an instance.

Shared memory region holding buffer caches, shared SQL areas, and other structures accessed by all server processes. Its size and component allocation strongly affect performance and are set at instance startup.

Private memory allocated per server process, holding session-specific data such as sort areas and cursor state. It is not shared between sessions and grows with connection count and work area demands.

A server-managed process performing instance-wide work such as writing dirty buffers, archiving redo, or cleaning up after failed processes. These run independently of user sessions and are essential to instance health.

A logical storage container that groups one or more data files. Schema objects are created inside tablespaces, letting administrators control placement, quotas, and backup scope separately from physical file layout.

An operating-system file belonging to exactly one tablespace that stores the actual blocks of database objects. Its size, autoextend settings, and placement determine capacity and I/O distribution.

A small binary file recording the database's physical structure, including data file locations, log sequence, and checkpoint information. The instance cannot mount the database without a valid copy.

A set of files recording every change made to database blocks, used to replay work after a failure. Logs are written sequentially and organized into groups and members for redundancy.

A background process that copies filled online redo log files to an archive destination before they are reused. Archived logs enable point-in-time recovery and are required in archive log mode.

An event that forces all dirty buffers up to a specific system change number to be written to disk and records that position. It bounds crash recovery time and advances the checkpoint in the control file.

A monotonically increasing logical clock stamped on changes and transactions. It provides a consistent ordering used for read consistency, recovery targets, and flashback operations.

A contiguous run of blocks allocated as a single unit inside a data file. Objects grow by acquiring extents, and extent sizing policy affects fragmentation and space waste.

A set of extents allocated to hold a single database object such as a table, index, or undo area. Segments may span multiple data files within one tablespace.

The smallest unit of storage the database reads and writes, fixed at a chosen size when the database is created. Block size influences row density, contention, and I/O efficiency.

A named collection of objects such as tables, indexes, and views owned by a single database user. Object names are unique within a schema, and access is controlled by privileges.

Storage holding before-image data used to roll back transactions and provide read-consistent views to other sessions. Undo retention settings determine how long past versions remain available.

The shared memory area holding copies of data blocks read from disk. Cache hit ratio and replacement behavior determine how often the instance must perform physical reads.

An I/O operation that fetches a block from disk into memory because it is not found in the buffer cache. High physical read rates often indicate cache pressure or inefficient access paths.

The ordered set of operations the optimizer chooses to retrieve a query's rows, including join methods, access paths, and estimated costs. It explains why a statement performs as it does.

The component that evaluates candidate access paths and join orders using statistics and cost estimates, then selects the plan for a statement. Its choices depend heavily on accurate object statistics.

Metadata describing table row counts, column value distributions, and index characteristics. The optimizer uses them to estimate cardinality, so stale or missing statistics cause poor plan choices.

A separate structure that provides fast lookup of table rows by key value, at the cost of extra storage and maintenance on data changes. It can also satisfy queries without touching the table.

A logical unit of work that either commits all of its changes or rolls them back entirely. It provides atomicity and isolation, and its changes become durable at commit.

The operation that makes a transaction's changes permanent and visible to other sessions. It writes redo to disk and releases locks, but dirty data blocks may be written later.

The operation that reverses uncommitted changes using undo data and releases held locks. It returns the database to the state before the transaction began.

The guarantee that a query sees data as of a single point in time, even while other sessions modify rows. It is implemented using undo data to reconstruct older block versions.

A named bundle of privileges that can be granted to users or other roles to simplify administration. Roles can be enabled or disabled per session and support password protection.

A specific right to perform an action, such as querying a table or creating a session. System privileges apply database-wide, while object privileges apply to a named object.

A named set of password and resource limits assigned to users, covering lifetime, reuse, and lockout rules. Profiles enforce consistent account policy across many accounts.

A record of security-relevant activity such as logons, privilege use, or object access. Auditing can target statements, privileges, or objects and is reviewed to detect misuse.

Restoring data files from backup and applying archived redo up to a chosen moment, producing a database state from the past. It requires archived logs and is used to undo errors or data loss.

A family of features that rewinds data or the database to an earlier state without restoring files, using undo or flashback logs. It is much faster than traditional recovery for logical mistakes.

A network process that accepts incoming client connection requests and hands them to a server process. Its configuration determines which services and instances clients can reach.

A logical name through which clients connect to one or more instances, decoupling connection strings from physical server names. Services support workload routing and planned maintenance.

A private work area used for operations such as ordering, grouping, and hash joins. When it is too small, operations spill to temporary storage and slow down noticeably.

A named reason a session is not consuming CPU, such as waiting for disk I/O, a lock, or a network message. Analyzing the top wait events is a primary tuning technique.

A mechanism that prevents conflicting concurrent access to the same data, such as two sessions updating one row. Excessive blocking appears as lock-related wait events.