A field required in every MongoDB document. The
_id field must have a unique value. You can
think of the _id field as the document's primary key.
If you create a new document without an _id field, MongoDB
automatically creates the field and assigns a unique
BSON ObjectId to the field.
absolute system CPU utilization
System CPU utilization relative to the full amount of CPU available
to cloud instances that share CPU.
When a cloud provider throttles CPU utilization for a cloud instance,
the instance's absolute system CPU utilization is equal to the
baseline CPU utilization assigned to this instance.
When a cloud provider adds CPU above the baseline CPU, such as
through a bursting mechanism, the sum of normalized kernel CPU
utilization and user CPU utilization on an instance can exceed
the instance's baseline CPU. In this case, the sum of the normalized
kernel CPU utilization and user CPU utilization is still less than
the full amount of CPU shared by cloud instances. See also
relative system CPU utilization, baseline CPU utilization,
and burstable instances.
An operation the user can perform on a resource. Actions and
resources combine to create privileges. See action.
admin database
A privileged database. Users
must have access to the admin database to run certain
administrative commands. For a list of administrative commands,
see Administration Commands.
Advanced Persistent Threat
In security, an attacker who gains and maintains long-term access to the
network, disk and/or memory and remains undetected for an extended
period.
aggregation
An operation that reduces and summarizes large
sets of data. MongoDB's aggregate() and
mapReduce() methods are two
examples of aggregation operations. For more information, see
Aggregation Operations.
aggregation pipeline
Consists of one or more stages that process documents. Aggregation
operators calculate aggregate values without having to use
map-reduce. For a list of operators, see
Aggregation Reference.
alert
Notification sent by Atlas when your database operations or
server usage reach thresholds that affect cluster
performance. To learn what conditions you can set to trigger
alerts, see Resolve Alerts
analytics node
Specialized read-only node that can isolate queries which you do
not want to affect your operational workload. Analytics nodes are
useful for handling analytic data such as reporting queries
executed by BI tools. You can host analytics nodes in dedicated
geographic regions to optimize read performance and reduce
latency.
API
Communication protocol facilitating interaction between the
client and MongoDB Atlas. You can use the
Atlas Administration API to
automate many of the tasks performed in the Atlas UI.
Computational technique used to quickly find points in a dataset
that are close to a given query point. Atlas Vector Search uses ANN search
to find vector embeddings in the data that are closest to the
vector embeddings in the query without scanning every
vector.
A replica set member that exists just to vote in
elections. Arbiters do not replicate data. An
arbiter participates in elections for a primary but cannot
become a primary. For more details, see
Replica Set Arbiter.
Atlas
MongoDB Atlas
is a cloud-hosted database-as-a-service.
Atlas Search
Fine-grained text indexing enabling advanced text search on your
data without any additional required management. Atlas Search provides
options for several kinds of text analyzers, score-based results ranking, and a rich
Atlas Search Overview
Atlas user
Account used to access the Atlas application. You can grant Atlas
users access to Atlas organizations, projects, or both, with certain
permissions defined by user roles. A Atlas user is
different than a database user. Atlas users
do not provide access to any MongoDB databases.
Feature in Atlas that allows you to perform semantic search
on vector embeddings by comparing query vectors with indexed
vectors to find the closest match.
atomic operation
An atomic operation is a write operation that either completes
entirely or doesn't complete at all. For
distributed transactions, which
involve writes to multiple documents, all writes to each document must
succeed for the transaction to succeed. Atomic operations cannot
partially complete. See Atomicity and Transactions.
When using In-Use Encryption, automatically performing
encryption and decryption based on your preconfigured
encryption schema. The Automatic Encryption Shared Library translates MongoDB
Query Language into the correct call, meaning you don't need to
rewrite your application for specific encrypt and decrypt calls.
B-tree
A data structure commonly used by database management systems to
store indexes. MongoDB uses B-tree indexes.
backup
Copy of your data that encapsulates the state of your cluster at
a given time. Backups provide a safety measure in the case of
data loss events.
A tailable cursor that points to a list of backup files.
Backup cursors are for internal use only.
balancer
An internal MongoDB process that runs in the context of a
sharded cluster and manages the migration of chunks. Administrators must disable the balancer for all
maintenance operations on a sharded cluster. See
Sharded Cluster Balancer.
baseline CPU utilization
Fraction of the full amount of CPU available to cloud instances that
share CPU. A cloud provider assigns a certain amount of baseline CPU
to each cloud instance, based on the instance's cluster tier.
Typically, baseline CPU utilization falls between 20% and 50% of
absolute system CPU utilization. See also relative system CPU utilization
and burstable instances.
big-endian
A byte order in which the most significant byte (big end) of a
multibyte data value is stored at the lowest memory address.
click to enlarge
blocking sort
A sort that must be performed in memory before the output is returned.
In-memory sorts may impact performance for large data sets. Use an
indexed sort to avoid an in-memory sort.
bounded collection scan
A plan used by the query optimizer that
excludes documents with specific field value ranges. For
example, if a range of date field values is outside of a specified
date range, the documents in that range are excluded from the
query plan. See Collection Scan.
BSON
A serialization format used to store documents and make
remote procedure calls in MongoDB. "BSON" is a combination of the words
"binary" and "JSON". Think of BSON as a binary representation
of JSON (JavaScript Object Notation) documents. See
BSON Types and
MongoDB Extended JSON (v2).
BSON types
The set of types supported by the BSON serialization
format. For a list of BSON types, see BSON Types.
Given three properties of computing systems, consistency,
availability, and partition tolerance, a distributed computing
system can provide any two of these features, but never all
three.
capped collection
A fixed-sized collection that automatically
overwrites its oldest entries when the collection reaches its maximum size.
The MongoDB oplog that is used in replication is a
capped collection. See Capped Collections.
cardinality
The measure of the number of elements within a set of values.
For example, the set A = { 2, 4, 6 } contains 3 elements,
and has a cardinality of 3. See Shard Key Cardinality.
cartesian product
The result of combining two data sets where the combined set
contains every possible combination of values.
cfq
Complete Fairness Queueing (cfq) is an I/O operation scheduler
that allocates bandwidth for incoming request processes.
checksum
A calculated value used to ensure data integrity.
The md5 algorithm is sometimes used as a checksum.
chunk
A contiguous range of shard key values within a
shard. Chunk ranges are inclusive of the lower boundary
and exclusive of the upper boundary. MongoDB splits chunks when
they grow bigger than the configured chunk size. The default chunk
size is 128 megabytes. MongoDB migrates chunks when a shard
contains too many chunks of a collection relative to other shards.
For more details, see Data Partitioning with Chunks,
Sharded Cluster Balancer, and Manage Sharded Cluster Balancer.
client
The application layer that uses a database for data persistence
and storage. Drivers provide the interface
level between the application layer and the database server.
A client can also be a single thread or process.
client affinity
A consistent client connection to a specified data source.
cloud backups
Localized cluster backup storage using the native
snapshot functionality of the cluster's cloud service
provider.
Atlas supports Cloud Backups for clusters served on:
Dictates the memory, storage, vCPUs, and IOPS specification for
each data-bearing server in the cluster. Cluster storage size and
overall performance increase as the cluster tier increases.
A grouping of MongoDB documents. A collection
is the equivalent of an RDBMS table. A collection is
in a single database. Collections do not enforce a
schema. Documents in a collection can have different fields.
Typically, documents in a collection have a similar or related
purpose. See Namespaces.
collection scan
Collection scans are a query execution strategy where MongoDB must
inspect every document in a collection to see if it matches the
query criteria. These queries are very inefficient and don't use
indexes. See Query Optimization for details about
query execution strategies.
commit
Saves data changes made after the start of the
startSession command. Operations within a
transaction are not permanent until they are committed
with the commitTransaction command.
commit quorum
During an index build
the commit quorum
specifies how many secondaries must be ready to commit their local
index build before the primary node performs the commit.
Concurrency control ensures that database operations can be
executed concurrently without compromising correctness.
Pessimistic concurrency control, such as that used in systems
with locks, blocks any potentially
conflicting operations even if they may not conflict.
Optimistic concurrency control, the approach
used by WiredTiger, delays
checking until after a conflict may have occurred, ending and
retrying one of the operations in any write conflict.
config database
An internal database with metadata for a sharded cluster.
Typically, you don't modify the config database. For more
information about the config database, see
Config Database.
A mongod instance that stores all the metadata
associated with a sharded cluster and can also store
application data. See
Config Shards.
connection pool
A cache of database connections maintained by the driver. The cached
connections are re-used when connections to the database are
required, instead of opening new connections.
connection storm
A scenario where a driver attempts to open more connections to a
deployment than that deployment can handle. When requests for new
connections fail, the driver requests to establish even more
connections in response to the deployment slowing down or failing
to open new connections. These continuous requests can overload
the deployment and lead to outages.
container
A collected set of software and its dependent libraries that are
packaged together to make transferring between computing
environments easier. Containers run as compartmentalized processes
on your operating system, and can be given their own resource
constraints. Common container technologies are Docker and
Kubernetes.
contention factor
Multiple operations attempting to modify the same resource, such as a
document field, cause conflicts that delay operations. The
contention factor is a setting used with Queryable Encryption to internally
partition encrypted field/value pairs and optimize operations.
See contention.
cosine similarity
Metric that uses the angle between two vectors to determine the
similarity between those vectors. Cosine similarity is sensitive
to vector orientation. You can use cosine similarity function when
indexing your vector embeddings for Atlas Vector Search. If the vectors are
normalized to unit length, use dotProduct similarity
function instead.
CPU steal
The percentage by which the CPU usage exceeds the guaranteed baseline CPU
credit accumulation rate. CPU steal is relevant for cloud providers that
rely on the credit model in their bursting strategy. CPU credits are units
of CPU utilization that you accumulate. The credits accumulate at a
constant rate to provide a guaranteed level of performance. You can use
these credits for additional CPU performance. When the credit balance is
exhausted, MongoDB only provides the guaranteed baseline of CPU
performance and displays the amount of excess as steal percent See also
relative system CPU utilization, baseline CPU utilization,
and burstable instances.
CRUD
An acronym for the fundamental operations of a database: Create,
Read, Update, and Delete. See MongoDB CRUD Operations.
CSV
A text data format with comma-separated values.
CSV files can be used to exchange data between relational
databases because CSV files have tabular data. You can
import CSV files using mongoimport.
cursor
A pointer to the result set of a query. Clients can
iterate through a cursor to retrieve results. By default, cursors
not opened within a session automatically timeout after 10
minutes of inactivity. Cursors opened in a session close with
the end or timeout of the session.
See Iterate a Cursor in mongosh.
custom role
Custom set of MongoDB privilege actions and MongoDB
roles that you can save and assign to a
database user. Create custom roles when
Atlas's built-in roles don't describe your desired set of
privileges.
A key that encrypts your Data Encryption Key.
The customer master key must be hosted in a remote key
provider.
daemon
A background, non-interactive process.
data directory
The file system location where mongod stores data
files. dbPath specifies the data directory.
Data Encryption Key
A key you use to encrypt the fields in your MongoDB
documents. The encrypted Data Encryption Key is stored in your
Key Vault collection. The Data Encryption Key is
encrypted by the Customer Master Key.
Data Explorer
Tool within Atlas to view and interact with cluster data.
You can also use the Data Explorer to manage indexes and run
aggregation pipelines to process your data.
MongoDB's solution for querying data stored in low-cost S3
buckets, Atlas clusters, and HTTP endpoints using the
MongoDB Query Language. Analytics applications can use Atlas Data Federation to
make use of archived data for their data processing needs.
A property that allows clients to address members in a system
based on their locations. Replica sets
implement data-center awareness using tagging. See
Data Center Awareness.
database
A container for collections.
Each database has a set of files in the file
system. One MongoDB server typically has multiple
databases.
database command
A MongoDB operation, other than an insert, update, remove, or
query. For a list of database commands, see
Database Commands. To use database commands, see
Issue Commands.
database exfiltration
Database exfiltration refers to an authorized party taking data from
a secured system and either sharing it with an unauthorized party or
storing it on an unsecured system. This may be malicious or accidental.
database profiler
A tool that, when enabled, keeps a record on all long-running
operations in a database's system.profile collection. The
profiler is most often used to diagnose slow queries. See
Database Profiler.
database user
Credentials used to authenticate a client to access a MongoDB
cluster. You can assign privileges to a database user to determine that
user's access level to a cluster. Database users
are different from Atlas users. Database
users have access to MongoDB deployments, not the Atlas
application.
A replica set member that cannot become primary and
applies operations at a specified delay. The delay is useful for
protecting data from human error (unintentionally deleted
databases) or updates that have unforeseen effects on the
production database. See Delayed Replica Set Members.
dense vectors
Numeric representation of data where most or all of the dimensions
contain non-zero values. Atlas Vector Search uses dense vectors, which are
packed with more data, to capture more complex relationships.
deployment
A group of MongoDB servers containing your data.
Atlas-managed clusters are clusters
(replica sets or
sharded clusters).
dimensions
Number of components or elements that make up the features or
attributes of data in multi-dimensional space. Atlas Vector Search supports up
to 4096 dimensions at index-time and query-time.
document
A record in a MongoDB collection and the basic unit of
data in MongoDB. Documents are analogous to JSON objects
but exist in the database in a more type-rich format known as
BSON. See Documents.
dot notation
MongoDB uses the dot notation to access the elements of an array
and to access the fields of an embedded document. See
Dot Notation.
dotProduct similarity
Measures similarity between two vectors in multi-dimensional
space and returns a scalar value, which is positive when the
vectors point in roughly the same direction, negative when the
vectors point in opposite direction, and zero when the vectors
have no similarity. Atlas Vector Search supports using dotproduct
similarity function when searching for nearest neighbors. We
recommend this similarity function instead of cosine similarity if
the vectors are normalized to unit length.
A client library for interacting with MongoDB in a particular
computer language. See driver.
durable
A write operation is durable when it persists after a shutdown (or
crash) and restart of one or more server processes. For a single
mongod server, a write operation is considered
durable when it has been written to the server's journal
file. For a replica set, a write operation
is considered durable after the write operation achieves
durability on a majority of voting nodes and written to a majority
of voting nodes' journals.
electable node
Node which is eligible to become the primary member of
your replica set. Atlas prioritizes nodes in the
highest priority region for primary eligibility during
elections. To ensure reliable elections, the total number of
electable nodes across an entire region must be 3, 5, or 7.
Representation of data such as text, images, audio, video,
and so on as an array of numbers, which can be interpreted as
coordinates in multi-dimensional space. Atlas supports storing
embeddings in an Atlas cluster and Atlas Vector Search supports indexing
and querying vector embeddings of up to 4096 dimensions.
encryption key
Random string of bits generated specifically to encrypt and
decrypt data.
Atlas Project Owners can
configure an additional layer of encryption on their data in
addition to the default encryption at rest that Atlas provides. Project
owners can use their Atlas-compatible customer key management
provider with the MongoDB encrypted storage engine.
Atlas supports the following customer key management providers
when configuring Encryption at Rest:
In JSON schema
that defines which fields are queryable and which query types are
permitted on those fields.
endianness
In computing, endianness refers to the order in which bytes are
arranged. This ordering can refer to transmission over a
communication medium or more commonly how the bytes are ordered
in computer memory, based on their significance and position.
For details, see big-endian and little-endian.
envelope encryption
An encryption procedure where data is encrypted using a
Data Encryption Key and the data encryption key is
encrypted by another key called the Customer Master Key.
The encrypted keys are stored as BSON documents in a
MongoDB collection called the KeyVault.
euclidean similarity
Formula to calculate the similarity by using the distance between
two vectors in multi-dimensional space. Euclidean distance is
sensitive to the magnitude of the vectors. Atlas Vector Search supports using
euclidean similarity function for indexing vectors and when
searching for nearest neighbors.
eventual consistency
A property of a distributed system that allows changes to the
system to propagate gradually. In a database system, this means
that readable members aren't required to have the latest
updates.
explicit encryption
When using In-Use Encryption, explicitly specifying the
encryption or decryption operation, keyID, and
query type (for Queryable Encryption) or algorithm (for Client-Side Field Level Encryption) when working
with encrypted data. Compare to automatic encryption.
expression
A component of a query that resolves to a value. Expressions are
stateless, meaning they return a value without mutating any of the
values used to build the expression.
In the MongoDB Query Language, you can build expressions from the
following components:
A name-value pair in a document. A document has
zero or more fields. Fields are analogous to columns in relational
databases. See Document Structure.
field path
Path to a field in a document. To specify a field path, use a
string that prefixes the field name with a dollar sign ($).
firewall
A system level network filter that restricts access based on
IP addresses and other parameters. Firewalls are part of a
secure network. See Firewalls.
free tier
Free-to-use cluster tier that provides a small-scale development
environment to host your data. Free clusters never expire,
and provide access to a subset of Atlas
features and functionality. Free clusters might also be
referred to by their instance size, M0.
Clusters with defined geographic zones to support location-aware
read and write operations for globally distributed application
instances and clients. You can enable global sharding on clusters
of tier M30 and greater.
The available geographic regions depend on the selected cloud
service provider.
GridFS
A convention for storing large files in a MongoDB database. All of
the official MongoDB drivers support the GridFS convention, as does the
mongofiles program. See GridFS for Self-Managed Deployments.
A health manager runs health checks on a health manager facet at a specified intensity level. The health manager checks are run at
specified time intervals. A health manager can be configured to
move a failing mongos out of a cluster
automatically.
health manager facet
A set of features that a health manager can be configured to run health checks for. For
example, you can configure a health manager to monitor and
manage DNS or LDAP cluster health issues automatically. See
Health Manager Facets for details.
Algorithm for performing efficient nearest neighbor search in
multi-dimensional space. Atlas Vector Search performs ANN search with
Hierarchical Navigable Small Worlds.
high availability
High availability indicates a system designed for durability,
redundancy, and automatic failover. Applications
supported by the system can operate without
downtime for a long time period. MongoDB
replica sets support
high availability when deployed according to the
best practices.
Method of combining different search methods, such as a full-text
and a semantic search, to take advantage of their respective
strengths. The results are combined by using a technique such as
Reciprocal Rank Fusion (RRF).
idempotent
An operation produces the same result with the
same input when run multiple times.
impact
Estimated performance improvement of an index that Performance Advisor suggests.
A sort that must be performed in memory before the output is
returned. In-memory sorts may impact performance for large data sets. Use an
indexed sort to avoid an in-memory sort.
Encryption that secures data when transmitted, stored, and
processed, and enables supported queries on that encrypted data.
MongoDB provides two approaches to In-Use Encryption:
Queryable Encryption and
Client-Side Field Level Encryption.
index
A data structure that optimizes queries. See Indexes.
index bounds
The range of index values that MongoDB searches when using an
index to run a query. To learn more, see
Multikey Index Bounds.
indexed sort
A sort where an index provides the sorted result. Sort operations that
use an index often have better performance than an in-memory sort.
See Use Indexed to Sort Query Results for
more information.
init script
A shell script used by a Linux platform's
init system to start, restart, or stop a daemon
process. If you installed MongoDB using a package manager, an init
script is provided for your system as part of the
installation. See the respective
Installation Guide for your operating
system.
init system
The init system is the first process started on a Linux platform
after the kernel starts, and manages all other processes on the
system. The init system uses an init script to start,
restart, or stop a daemon process, such as
mongod or mongos. Recent Linux
versions typically use the systemd init system and the
systemctl command. Older Linux versions typically use the
System V init system and the service command. See
the Installation Guide for your operating system.
initial sync
The replica set operation that replicates data from an
existing replica set member to a new replica set member. See
Initial Sync.
intent lock
A lock on a resource that indicates the lock holder
will read from (intent shared) or write to (intent
exclusive) the resource using concurrency control at
a finer granularity than that of the resource with the intent
lock. Intent locks allow concurrent readers and writers of a
resource. See What type of locking does MongoDB use?.
interface endpoint
AWSVPC endpoint with a private IP address
that sends traffic to the Atlas private endpoint service over AWS PrivateLink.
A point in an operation when it can
safely end. MongoDB only ends an operation
at designated interrupt points. See
Terminate Running Operations.
IP access list
List of IP addresses and CIDR blocks with access to clusters
within an Atlas project. For client connections over the
public Internet, Atlas allows connections to a cluster only
from entries in the corresponding project's IP access list.
The access list may have up to 200 entries.
A revision to the IP (Internet Protocol) standard with a
large address space to support Internet hosts.
ISODate
The international date format used by mongosh
to display dates. The format is YYYY-MM-DD HH:MM.SS.millis.
JavaScript
A scripting language. mongosh, the legacy
mongo shell, and certain server
functions use a JavaScript interpreter. See
Server-side JavaScript for more information.
journal
A sequential, binary transaction log used to bring the database
into a valid state in the event of a hard shutdown.
Journaling writes data first to the journal and then to the core
data files. MongoDB enables journaling by default for 64-bit
builds of MongoDB version 2.0 and newer. Journal files are
pre-allocated and exist as files in the data directory.
See Journaling.
JSON
JavaScript Object Notation. A plain text format
for expressing structured data with support in many programming
languages. For more information, see http://www.json.org.
Certain MongoDB tools render an approximation of MongoDB
BSON documents in JSON format. See
MongoDB Extended JSON (v2).
JSON document
A JSON document is a collection of fields and values in a
structured format. For sample JSON documents, see
http://json.org/example.html.
JSONP
JSON with padding. Refers to a method of injecting JSON
into applications. Presents potential security concerns.
Given a set of points P with a defined similarity function S,
for a query point q, finds the set of k points in P with the
best values of S*(*p, q). Atlas Vector Search ENN search returns the
exact top k points and ANN search returns k points that are
similar to q, but not necessarily the k most similar to q.
key material
The random string of bits used by an encryption algorithm to
encrypt and decrypt data.
Cross-platform protocol used to authenticate users and authorize
them to access data on a cluster. You can use Atlas to manage
user authentication and authorization from all MongoDB clients
using your own LDAP server over TLS. A single LDAPS
configuration applies to all clusters in an Atlas project.
least privilege
An authorization policy that grants a user only the access
that is essential to that user's work.
legacy coordinate pairs
The format used for geospatial data before MongoDB
version 2.4. This format stores geospatial data as points on a
planar coordinate system (for example, [ x, y ]). See
Geospatial Queries.
LineString
A LineString is an array of two or more positions. A
closed LineString with four or more positions is called a
LinearRing, as described in the GeoJSON LineString specification:
https://tools.ietf.org/html/rfc7946#section-3.1.4. To use a
LineString in MongoDB, see
GeoJSON Objects.
link-token
String that contains the information necessary to connect from
Cloud Manager or Ops Manager to Atlas during a live migration from a Cloud Manager or Ops Manager deployment
to a cluster in Atlas.
When you are ready to live migrate data from a Cloud Manager or Ops Manager deployment,
you generate a link-token in Atlas and then enter it in your
Cloud Manager or Ops Manager organization's settings. You use the same link-token to
migrate each deployment in your Cloud Manager or Ops Manager organization sequentially,
one at a time. You can generate multiple link-tokens in Atlas.
Use one unique link-token for each Cloud Manager or Ops Manager organization.
little-endian
A byte order in which the least significant byte (little end)
of a multibyte data value is stored at the lowest memory address.
click to enlarge
live migration
Process to seamlessly move an existing source replica set or
sharded cluster to Atlas. During the live migration process,
Atlas keeps the target cluster in sync with the remote source
until you cut your applications over to the Atlas cluster.
Atlas offers two modes of live migration:
Push live migration, known in the Atlas user interface as
Live Migration from Ops Manager or Cloud Manager,
where Atlas pushes a deployment from Cloud Manager or Ops Manager to Atlas.
Pull live migration, known in the Atlas user interface as
General Live Migration, where Atlas pulls
a deployment from a cloud or on-premise deployment to Atlas.
Contain server events, such as incoming connections, commands run,
and issues encountered. For more details, see
Log Messages.
LVM
Logical volume manager. LVM is a program that abstracts disk
images from physical devices and provides a number of raw disk
manipulation and snapshot capabilities useful for system
management. For information on LVM and MongoDB, see
Back Up and Restore Using LVM on Linux.
maintenance window
Day and time of the week when Atlas should start weekly
maintenance on your cluster. You can set your maintenance window
in your Project Settings.
Important
Maintenance Window Considerations
Urgent Maintenance Activities
Urgent maintenance activities such as security patches cannot
wait for your chosen window. Atlas will start those
maintenance activities when needed.
Ongoing Maintenance Operations
Once maintenance is scheduled for your cluster, you cannot change
your maintenance window until the current maintenance efforts have
completed.
Maintenance Requires Replica Set Elections
Atlas performs maintenance the same way as the maintenance
procedure described in the
MongoDB Manual. This
procedure requires at least one
replica set election
during the maintenance window per replica set.
Maintenance Starts As Close to the Hour As Possible
Maintenance always begins as close to the scheduled hour as
possible, but in-progress cluster updates or unexpected system
issues could delay the start time.
map-reduce
An aggregation process that has a "map"
phase that selects the data and a "reduce" phase that transforms the
data. In MongoDB, you can run arbitrary aggregations over data
using map-reduce. For the map-reduce implementation, see
Map-Reduce. For all approaches to aggregation,
see Aggregation Operations.
mapping type
A structure in programming languages that associate keys with
values. Keys may contain embedded pairs of keys and values
(for example, dictionaries, hashes, maps, and associative arrays).
The properties of these structures depend on the language
specification and implementation. Typically, the order of keys in
mapping types is arbitrary and not guaranteed.
md5
A hashing algorithm that calculates a checksum for the
supplied data. The algorithm returns a unique value
to identify the data. MongoDB uses md5 to identify chunks of data
for GridFS. See filemd5.
mean
Average of a set of numbers.
median
In a dataset, the median is the percentile value where 50% of the
data falls at or below that value.
member
An individual mongod process. A replica set has
multiple members. A member is also known as a node.
Multipurpose Internet Mail Extensions. A standard set of type and
encoding definitions used to declare the encoding and type of data
in multiple data storage, transmission, and email contexts. The
mongofiles tool provides an option to specify a MIME
type to describe a file inserted into GridFS storage.
mode
Number that occurs most frequently in a set of numbers.
mongo
The legacy MongoDB shell. The mongo process starts
the legacy shell as a daemon connected to either a
mongod or mongos instance. The shell
has a JavaScript interface.
Starting in MongoDB v5.0, mongo is deprecated and
mongosh replaces mongo as the
client shell. See mongosh.
mongod
The MongoDB database server. The mongod process
starts the MongoDB server as a daemon. The MongoDB server
manages data requests and background operations. See
mongod.
MongoDB Charts
Visualization tool for your Atlas data. You can launch MongoDB
Charts from your Atlas cluster and view your data with the
Charts application to begin visualizing your data.
The MongoDB sharded cluster query router. The
mongos process starts the MongoDB router as a
daemon. The MongoDB router acts as an interface
between an application and a MongoDB sharded cluster and
handles all routing and load balancing across the cluster. See
mongos Instances.
mongosh
MongoDB Shell. mongosh provides a shell
interface to either a mongod or a
mongos instance.
Starting in MongoDB v5.0, mongosh replaces
mongo as the preferred shell.
multi-region cluster
Atlas cluster spanning multiple geographic regions.
Multi-region clusters can increase availability and improve
performance by routing application queries to the most appropriate
geographic regions.
A namespace is a combination of the database name and
the name of the collection or index:
<database-name>.<collection-or-index-name>. All documents
belong to a namespace. See Namespaces.
Namespace Insights
Atlas tool that monitors collection-level
query latency. You can view query latency metrics and
statistics for certain hosts and operation types. Manage pinned
namespaces and choose up to five namespaces to show in the
corresponding query latency charts.
The order recordIds are created and stored in the
WiredTiger index. The default sort
order for a collection scan run on a single instance is
natural order.
In replica sets, natural order is not guaranteed to be consistent
and can differ between members.
In sharded collections, natural order is not defined. However,
using $natural still forces each shard to perform a
collection scan.
A network failure that separates a distributed system into
partitions such that nodes in one partition cannot communicate
with the nodes in the other partition.
Sometimes, partitions are partial or asymmetric. An example
partial partition is the a division of the nodes of a network
into three sets, where members of the first set cannot
communicate with members of the second set, and the reverse, but
all nodes can communicate with members of the third set.
In an
asymmetric partition, communication may be possible only when it
originates with certain nodes. For example, nodes on one side of
the partition can communicate with the other side only if they
originate the communications channel.
network peering connection
Process by which two Internet networks connect and exchange
traffic. You can directly peer your VPC with the Atlas
VPC created for your MongoDB clusters. Using network peering,
your application servers can directly connect to Atlas while
remaining isolated from public networks.
An individual mongod process. A replica set has
multiple nodes. A node is also known as a member.
noop
No Operation (noop), is an I/O operation scheduler that allocates
I/O bandwidth for incoming processes based on a first in, first out
queue.
NVMe
NVMe (Non-Volatile Memory Express) is a protocol for accessing
high-speed storage media.
NVMe storage
Available for M40+ clusters hosted on AWS
For applications hosted on AWS which require low-latency and
high-throughput IO, you can use the NVMe cluster class.
The NVMe cluster class leverages a unique data protocol to
greatly improve data access speeds.
NVMe clusters use a hidden secondary node consisting of a provisioned
volume with high throughput and IOPS to facilitate backup.
A 12-byte BSON type that is unique
within a collection. The ObjectId is generated using the
timestamp, computer ID, process ID, and a local process incremental
counter. MongoDB uses ObjectId values as the default values for
_id fields.
A keyword beginning with a $ used to express an update,
complex query, or data transformation. For example, $gt is the
query language's "greater than" operator. For available operators,
see Operators.
A temporary collection created during resharding operations that stores oplog entries from a
donor shard.
Oplog buffer collections ensure that recipient shards
can access oplog entries when they get deleted from the donor shard.
Oplog buffer collections are removed when resharding is complete.
oplog hole
A temporary gap in the oplog because the oplog writes aren't in
sequence. Replica set primaries apply
oplog entries in parallel as a batch operation. As a result,
temporary gaps in the oplog can occur from entries that aren't
yet written from a batch.
oplog window
oplog entries are time-stamped. The oplog window is the time
difference between the newest and the oldest timestamps in the
oplog. If a secondary node loses connection with the primary, it
can only use replication to sync up again if the
connection is restored within the oplog window.
optime
A reference to a position in the replication oplog. The optime
value is a document that contains:
Unique 24-digit hexadecimal string used to identify your
Atlas organization. The Return All
Organizations endpoint returns the ID
of all organizations that the authenticated user executing the
API call can access.
orphaned cursor
A cursor that is not correctly closed or iterated over
in your application code. Orphaned cursors can cause performance
issues in your MongoDB deployment.
orphaned document
In a sharded cluster, orphaned documents are those documents on a
shard that also exist in chunks on other shards. This is caused by
a failed migration or an incomplete migration cleanup because of
an atypical shutdown.
Orphaned documents are cleaned up automatically after a chunk migration
completes. You no longer need to run cleanupOrphaned to
delete orphaned documents.
A type of cache that locally stores memory for a specific CPU core.
Per-CPU caches are used by the new version of TCMalloc, which is
introduced in MongoDB 8.0.
per-thread cache
A type of cache that locally stores memory for each application thread.
Per-thread caches are used by the legacy version of TCMalloc, which is
used in MongoDB 7.0 and earlier.
percentile
In a dataset, a percentile is a value where that percentage
of the data is at or below the specified value. For details, see
Calculation Considerations.
Performance Advisor
Atlas tool that monitors slow queries executed on your
cluster and suggests indexes to improve query performance. Each
index that the Performance Advisor suggests include an
impact score indicating the potential performance
improvement that index would bring.
A process identifier. UNIX-like systems assign a unique-integer
PID to each running process. You can use a PID to inspect a
running process and send signals to it. See
/proc File System.
pipe
A communication channel in UNIX-like systems allowing independent
processes to send and receive data. In the UNIX shell, piped
operations allow users to direct the output of one command into
the input of another.
A combination of query predicate, sort, projection, and collation. The plan cache
query shape allows MongoDB to identify equivalent queries and
analyze their performance.
For the query predicate, only the predicate structure and field
names are used. The values in the query predicate aren't used. For
example, a query predicate { type: 'food' } is equivalent to
{ type: 'drink' }.
To identify slow queries with the same plan cache query shape,
each plan cache query shape has a hexadecimal planCacheShapeHash
value. For more information, see planCacheShapeHash and planCacheKey.
Starting in MongoDB 8.0, the existing queryHash field is duplicated
in a new field named planCacheShapeHash. If you're using an earlier
MongoDB version, you'll only see the queryHash field. Future MongoDB
versions will remove the deprecated queryHash field, and you'll need
to use the planCacheShapeHash field instead.
An array of LinearRing coordinate arrays, as
described in the GeoJSON Polygon specification:
https://tools.ietf.org/html/rfc7946#section-3.1.6. For Polygons
with multiple rings, the first must be the exterior ring and
any others must be interior rings or holes.
MongoDB does not permit the exterior ring to self-intersect.
Interior rings must be fully contained within the outer loop and
cannot intersect or overlap with each other. See
GeoJSON Objects.
A setting for each collection that allocates space for each
document to maximize storage reuse and reduce
fragmentation. powerOf2Sizes is the default for TTL
Collections. To change collection settings, see
collMod.
An operation performed before inserting data that divides the
range of possible shard key values into chunks to facilitate easy
insertion and high write throughput. In some cases pre-splitting
expedites the initial distribution of documents in sharded cluster by manually dividing the collection rather than waiting
for the MongoDB balancer to do so. See
Create Ranges in a Sharded Cluster.
prefix compression
Reduces memory and disk consumption by storing any identical index
key prefixes only once, per page of memory. See:
Compression for more about WiredTiger's
compression behavior.
primary
In a replica set, the primary is the member that
receives all write operations. See
Primary.
primary key
A record's unique immutable identifier. In RDBMS software, the primary
key is typically an integer stored in each row's id field.
In MongoDB, the _id field stores a document's primary
key, which is typically a BSON ObjectId.
primary shard
Each database in a sharded cluster has a primary shard. It is the
default shard for all unsharded collections in the database. See
Primary Shard.
Unique 24-digit hexadecimal string used to identify your
Atlas project. The Get All ProjectsAPI endpoint returns the ID of all
projects that the authenticated user executing the API call can
access.
Method of compressing the value of individual dimensions in a
vector into a smaller range to reduce resource consumption and
improve speed. Atlas Vector Search supports indexing and querying quantized
vectors.
A combination of the query optimizer and query execution engine
that processes an operation.
query operator
A keyword beginning with $ in a query. For example,
$gt is the "greater than" operator. For a list of
query operators, see query operators.
query optimizer
A process that generates query plans. For each query, the
optimizer generates a plan that matches the query to the index
that returns the results as efficiently as possible. The
optimizer reuses the query plan each time the query runs. If a
collection changes significantly, the optimizer creates a new
query plan. See Query Plans.
query plan
Most efficient execution plan chosen by the query planner. For
more details, see Query Plans.
query predicate
An expression that returns a boolean indicating whether a document
matches the specified query. For example, { name: { $eq: "Alice"
} }, which returns documents that have a field "name" whose
value is the string "Alice".
Query predicates can contain child expressions and operators for
more complex matching. To see available query operators, see
Query and Projection Operators.
Query Profiler
Atlas tool that diagnoses and monitors performance
issues in your cluster. The Query Profiler can expose long-running
queries and their performance statistics. You can filter the data
returned by the Query Profiler to hone in on specific namespaces
and operation types.
query shape
A query shape is a set of specifications that group similar queries.
For details, see Query Shapes.
Relational Database Management System. A database management
system based on the relational model, typically using SQL
as the query language.
read concern
Specifies a level of isolation for read operations. For example,
you can use read concern to only read data that has propagated to
a majority of nodes in a replica set. See
Read Concern.
A setting that determines how clients direct read operations.
Read preference affects all replica sets, including shard replica
sets. By default, MongoDB directs reads to primaries. However, you may also direct reads to secondaries for
eventually consistent reads. See
Read Preference.
read-only node
Replica set in a dedicated geographic region that supplements your
electable node regions. You can use read-only nodes to
localize data where it is most frequently read to improve
performance.
Real-Time Performance Panel
Atlas monitoring service that displays current network
traffic, database operations on your clusters, and hardware
statistics about your host machines. Use the RTPP to visually
evaluate query execution times, monitor network activity, and
discover potential replication lag on secondary members of
replica sets.
Measures the fraction of true nearest neighbors that were returned
by an ANN search. This measure reflects how close the algorithm
approximates the results of ENN search. The notation Recall@k
refers to the measurement of how many of the true nearest
neighbors were present in the top k results returned by Atlas Vector Search.
recovering
A replica set member status indicating that a member
is not ready to begin activities of a secondary or primary.
Recovering members are unavailable for reads.
relative system CPU utilization
The CPU utilization relative to the amount of baseline CPU assigned to
a cloud instance. You can calculate relative system CPU utilization
by dividing the absolute system CPU utilization by the amount
of baseline CPU assigned to a cloud instance.
MongoDB caps relative system CPU utilization at 100%. When a cloud provider
throttles CPU utilization for a cloud instance, or bursts CPU utilization
for an instance above the baseline amount of CPU available to that
instance, the relative system CPU value is 100%.
Group of MongoDB servers that maintain the same data set. Replica
sets provide redundancy, high availability, and are the basis for
all production deployments.
replication
A feature allowing multiple database servers to share the same
data. Replication ensures data redundancy and enables load
balancing. See Replication.
replication lag
The time period between the last operation in the
primary'soplog and the last operation
applied to a particular secondary. You typically want
replication lag as short as possible. See Replication
Lag.
resident memory
The subset of an application's memory currently stored in
physical RAM. Resident memory is a subset of virtual memory,
which includes memory mapped to physical RAM and to storage.
resource
A database, collection, set of collections, or cluster. A
privilege permits actions on a specified
resource. See resource.
role
A set of privileges that permit actions on
specified resources. Roles assigned to a user
determine the user's access to resources and operations. See
Security.
Process that restarts all nodes in the cluster in sequence. To
maintain cluster availability, Atlas restarts one node at a
time starting with a secondary node.
Atlas always maintains a primary node until the rolling
restart completes.
scalar quantization
Scalar quantization involves selecting the minimum and maximum values
across all indexed vectors within a segment for each dimension,
and producing equally sized bins between them. The mappings for
each of these dimensions to the bins yields the new quantized
values. Atlas Vector Search supports automatic scalar quantization for your
float32 vectors, and ingestion and indexing of your scalar
quantized vectors from embedding providers.
secondary
A replica set member that replicates the contents of the
master database. Secondary members may run read requests, but
only the primary members can run write operations. See
Secondaries.
secondary index
A database index that improves query performance by
minimizing the amount of work that the query engine must perform
to run a query. See Indexes.
A seed list is used by drivers and clients (like
mongosh) for initial discovery of the replica
set configuration. Seed lists can be provided as a list of
host:port pairs (see Standard Connection String Format
or through DNS entries.) For more information,
see SRV Connection Format.
self-managed
A MongoDB instance that is set up and maintained by an
individual or organization, and not an external management or
third-party services (such as MongoDB Atlas).
semantic search
Search for values that have a similar meaning to query. Semantic
search captures the natural relationship between words or phrases
even when there is no lexical overlap. Semantic search and vector
search are often used interchangeably. Atlas Vector Search supports semantic
search on vector data stored in Atlas clusters.
set name
The arbitrary name given to a replica set. All members of a
replica set must have the same name specified with the
replSetName setting or the --replSet option.
shard
A single mongod instance or replica set
that stores part of a sharded cluster's
total data set. Typically, in a production deployment, ensure all
shards are part of replica sets. See
Shards.
The set of nodes comprising a sharded MongoDB
deployment. A sharded cluster consists of config servers,
shards, and one or more mongos
routing processes. See Sharded Cluster Components.
sharding
A database architecture that partitions data by key ranges and
distributes the data among two or more database instances.
Sharding enables horizontal scaling. See Sharding.
shared cluster
Cluster category containing M0 (free tier) tier clusters.
Shared clusters are generally used for development and small
production workloads.
A method in mongosh that has a concise
syntax for a database command. Shell helpers
improve the interactive experience. See
mongosh Methods.
similarity function
Measures the similarity between two vectors. Atlas Vector Search supports
euclidean, cosine, and dotProduct similarity functions.
single-master replication
A replication topology where only a single database
instance accepts writes. Single-master replication ensures
consistency and is the replication topology used by MongoDB.
See Replica Set Primary.
snappy
A compression/decompression library to balance
efficient computation requirements with reasonable compression rates.
Snappy is the default compression
library for MongoDB's use of WiredTiger. See WiredTiger compression
documentation
for more information.
snapshot
A snapshot is a copy of the data in a mongod instance at a
specific point in time. You can retrieve snapshot metadata for the whole cluster
or replica set, or for a single config server in a cluster.
softIRQ
The CPU utilization metric that reflects a portion of CPU that a cloud
instance currently uses to process software interrupt requests.
On some cloud providers, this metric is useful for tracking CPU
utilization on burstable instances.
sort key
The value compared against when sorting fields. To learn how
MongoDB determines the sort key for non-numeric fields, see
Comparison/Sort Order.
Structured Query Language (SQL) is used for interaction with
relational databases.
SSD
Solid State Disk. High-performance storage that uses solid
state electronics for persistence instead of rotating platters
and movable read/write heads used by mechanical hard drives.
stale read
A stale read refers to when a transaction reads old (stale) data that has
been modified by another transaction but not yet committed to the
database.
A standalone instance is not a replica set with only one
member.
stash collection
A temporary collection created on the recipient shard for
each donor shard during resharding operations.
Stash collections temporarily hold documents that cannot be
immediately inserted due to operation conflicts. For
example, if a document's shard key has been updated, it now belongs
to a different shard, and the order of operations applied to this document
can be ambiguous. The recipient stores these documents in a
stash collection until it can apply operations
in the correct order.
step down
The primary member of the replica set removes
itself as primary and becomes a secondary member.
If a replica set loses contact with the primary, the
secondaries elect a new primary. When the old primary
learns of the election, it steps down and rejoins the
replica set as a secondary.
If the user runs the replSetStepDown
command, the primary steps down, forcing the replica set
to elect a new primary.
storage engine
The part of a database that is responsible for managing how data
is stored and accessed, both in memory and on disk. Different
storage engines perform better for specific workloads. See
Storage Engines for Self-Managed Deployments for specific details on the built-in
storage engines in MongoDB.
A property of a distributed system requiring that all members
contain the latest changes to the system. In a database
system, this means that any system that can provide data must
contain the latest writes.
Subject Alternative Name
Subject Alternative Name (SAN) is an extension of the X.509
certificate which allows an array of values such as IP addresses
and domain names that specify the resources a single security
certificate may secure.
sync
The replica set operation where members replicate data
from the primary. Sync first occurs when MongoDB creates
or restores a member, which is called initial sync. Sync
then occurs continually to keep the member updated with changes to
the replica set's data. See Replica Set Data Synchronization.
syslog
On UNIX-like systems, a logging process that provides a uniform
standard for servers and processes to submit logging information.
MongoDB provides an option to send output to the host's syslog
system. See syslogFacility.
tag
A label applied to a replica set member and used by
clients to issue data-center-aware operations. For more information
on using tags with replica sets, see Read Preference Tag Set Lists.
For a capped collection, a tailable cursor is a cursor that
remains open after the client exhausts the results in the initial
cursor. As clients insert new documents into the capped collection,
the tailable cursor continues to retrieve documents.
team
Group of Atlas users in the same organization. You can use teams to grant access to
the same group of Atlas users across multiple
projects. All users in the team share the same
project access.
term
For the members of a replica set, a monotonically increasing
number that corresponds to an election attempt.
time series collection
A collection that efficiently stores
sequences of measurements over a period of time. See
Time Series.
topology
The state of a deployment of MongoDB instances. Includes:
Type of deployment (standalone, replica set, or sharded cluster).
Group of read or write operations. For details, see
Transactions.
transaction coordinator
A component of MongoDB that manages transactions in a replica set or a sharded cluster. It coordinates the execution and completion of
multi-document transactions across nodes and allows a complex
operation to be treated as an atomic operation.
TSV
A text-based data format consisting of tab-separated values.
This format is commonly used to exchange data between relational
databases because the format is suited to tabular data. You can
import TSV files using mongoimport.
TTL
Time-to-live (TTL) is an expiration time or
period for a given piece of information to remain in a cache or
other temporary storage before the system deletes it or ages it
out. MongoDB has a TTL collection feature. See
Expire Data from Collections by Setting TTL.
unbounded array
An array that consistently grows larger over time. If a document
field value is an unbounded array, the array may negatively impact
performance. In general, design your schema to avoid unbounded
arrays.
unique index
An index that enforces uniqueness for a particular field in
a single collection. See Unique Indexes.
unix epoch
January 1st, 1970 at 00:00:00 UTC. Commonly used in expressing time,
where the number of seconds or milliseconds since this point is counted.
unordered query plan
A query plan that returns results in an order inconsistent with the
sort() order.
See Query Plans.
System that stores vector embeddings and associated metadata, and
enables nearest neighbor search on the stored vector embeddings.
You can use Atlas as your vector database and Atlas Vector Search to
perform vector search on the stored vector embeddings. You can use
vector database to implement RAG.
vector index
Data structure that efficiently processes nearest neighbor search
queries. Atlas Vector Search supports creating indexes of type vector to
index fields for running $vectorSearch queries.
vector search
Method of performing k nearest neighbor search over a set of
vectors stored in a vector index. Atlas Vector Search supports ANN and ENN
search for k nearest neighbors.
virtual memory
An application's working memory, typically residing on both
disk and in physical RAM.
WGS84
The default reference system and geodetic datum that MongoDB uses
to calculate geometry over an Earth-like sphere for geospatial
queries on GeoJSON objects. See the
"EPSG:4326: WGS 84" specification:
http://spatialreference.org/ref/epsg/4326/.
window operator
Returns values from a span of documents from a collection. See
window operators.
working set
The data that MongoDB uses most often.
write concern
Specifies whether a write operation has succeeded. Write concern
allows your application to detect insertion errors or unavailable
mongod instances. For replica sets, you can configure write concern to confirm replication to a
specified number of members. See Write Concern.
write conflict
A situation where two concurrent operations, at least one of which
is a write, try to use a resource that violates the
constraints for a storage engine that uses optimistic
concurrency control. MongoDB automatically ends and
retries one of the conflicting write operations.
write lock
An exclusive lock on a resource such as a collection
or database. When a process writes to a resource, it takes
an exclusive write lock to prevent other processes from writing
to or reading from that resource. For more information on
locks, see FAQ: Concurrency.
zlib
A data compression library that provides higher compression rates
at the cost of more CPU, compared to MongoDB's use of
snappy. You can configure WiredTiger to use zlib as its compression library. See
WiredTiger compression documentation for more
information.
zone
A grouping of documents based on ranges of shard key values
for a given sharded collection. Each shard in the sharded cluster can
be in one or more zones. In a balanced cluster, MongoDB
directs reads and writes for a zone only to those shards
inside that zone. See the Zones manual page for more
information.
zstd
A data compression library that provides higher compression rates
and lower CPU usage when compared to zlib.
