Memory configuration

This page describes the different aspects of Neo4j memory configuration and use. The RAM of the Neo4j server has a number of usage areas, with some sub-areas:

Figure 1. Neo4j memory management

OS memory

Some memory must be reserved for running the processes of the operating system itself. It is not possible to explicitly configure the amount of RAM that should be reserved for the operating system, as this is what RAM remains available after configuring Neo4j.

1GB is a good starting point for a server that is dedicated to running Neo4j. However, there are cases where the amount reserved for the OS is significantly larger than 1GB, such as servers with exceptionally large RAM.

If you have a vector index, you need to ensure that the OS has sufficient memory set aside for the vector index to perform optimally, because the vector index is loaded in OS memory and not in Neo4j page cache. For more information, see Vector index memory configuration.

If you do not leave enough space for the OS, it will start to swap memory to disk, which will heavily affect performance. Therefore, it is generally recommended to have swap turned off for a Neo4j dedicated server.

JVM Heap

The JVM heap is a separate dynamic memory allocation that Neo4j uses to store instantiated Java objects. The memory for the Java objects are managed automatically by a garbage collector. Particularly important is that a garbage collector automatically handles the deletion of unused objects. For more information on how the garbage collector works and how to tune it, see Tuning of the garbage collector.

The heap memory size is determined by the parameters server.memory.heap.initial_size and server.memory.heap.max_size. It is recommended to set these two parameters to the same value to avoid unwanted full garbage collection pauses.

Generally, to aid performance, you should configure a large enough heap to sustain concurrent operations.

Native memory

Native memory, sometimes referred to as off-heap memory, is memory directly allocated by Neo4j from the OS. This memory will grow dynamically as needed and is not subject to the garbage collector.

DBMS

The database management system, or DBMS, contains the global components of the Neo4j instance. For example, the bolt server, logging service, monitoring service, etc.

Database

Each database in the system comes with an overhead. In deployments with multiple databases, this overhead needs to be accounted for.

Transaction

When executing a transaction, Neo4j holds not yet committed data, the result, and intermediate states of the queries in memory. The size needed for this is very dependent on the nature of the usage of Neo4j. For example, long-running queries, or very complicated queries, are likely to require more memory. Some parts of the transactions can optionally be placed off-heap, but for the best performance, it is recommended to keep the default with everything on-heap.

This memory group can be limited with the setting dbms.memory.transaction.total.max.

Page cache

The page cache is used to cache the Neo4j data stored on disk. The caching of graph data and indexes into memory helps avoid costly disk access and result in optimal performance.

The parameter for specifying how much memory Neo4j is allowed to use for the page cache is: server.memory.pagecache.size.

Network buffers

Direct buffers are used by Neo4j to send and receive data. Direct byte buffers are important for improving performance because they allow native code and Java code to share data without copying it. However, they are expensive to create, which means byte buffers are usually reused once they are created.

Other shared buffers

This includes unspecified shared direct buffers.

JVM overhead

The JVM will require some memory to function correctly. For example, this can be:

Thread stacks – Each thread has its own call stack. The stack stores primitive local variables and object references along with the call stack (list of method invocations) itself. The stack is cleaned up as stack frames move out of context, so there is no GC performed here.
Metaspace – Metaspace stores the java class definitions and some other metadata.
Code cache – The JIT compiler stores the native code it generates in the code cache to improve performance by reusing it.

For more details and means of limiting the memory used by the JVM please consult your JVM documentation.

Considerations

Always use explicit configuration: To have good control of the system behavior, it is recommended to always define the page cache and heap size parameters explicitly in neo4j.conf. Otherwise, Neo4j computes some heuristic values at startup based on the available system resources.

Initial memory recommendation: Use the neo4j-admin server memory-recommendation command to get an initial recommendation for how to distribute a certain amount of memory. The values may need to be adjusted to cater for each specific use case.

Inspect the memory settings of all databases in a DBMS

The neo4j-admin server memory-recommendation command is useful for inspecting the current distribution of data and indexes.

Example 1. Use neo4j-admin server memory-recommendation to inspect the memory settings of all your databases

Estimate the total size of the database files.

bin/neo4j-admin server memory-recommendation
...
...
...
# Total size of lucene indexes in all databases: 6690m
# Total size of data and native indexes in all databases: 17050m

You can see that the Lucene indexes take up approximately 6.7GB of data, and that the data volume and native indexes combined take up approximately 17GB.

Using this information, you can do a sanity check of your memory configuration:

Compare the value for data volume and native indexes to the value of server.memory.pagecache.size.
For cases when off-heap transaction state is used, estimate transactional workload and how much memory is left to the value of dbms.tx_state.max_off_heap_memory.
Compare the value for Lucene indexes to how much memory is left after assigning server.memory.pagecache.size and server.memory.heap.initial_size.

In some production systems the access to memory is limited and must be negotiated between different areas. Therefore, it is recommended to perform a certain amount of testing and tuning of these settings to figure out the optimal division of the available memory.

Limit transaction memory usage recommendation: The measured heap usage of all transactions is only an estimate and the actual heap utilization may be slightly larger or slightly smaller than the estimated value. In some cases, limitations of the estimation algorithm to detect shared objects at a deeper level of the memory graph could lead to overestimations. This is because a conservative estimate is given based on aggregated estimations of memory usage, where the identities of all contributing objects are not known, and cannot be assumed to be shared. For example, when you use UNWIND on a very large list, or expand a variable length or shortest path pattern, where many relationships are shared between the computed result paths.

In these cases, if you experience problems with a query that gets terminated, you can execute the same query with the transaction memory limit disabled. If the actual heap usage is not too large, it might succeed without triggering an out-of-memory error.

Capacity planning

In many use cases, it is advantageous to try to cache as much of the data and indexes as possible. The following examples illustrate methods for estimating the page cache size, depending on whether you are already running in production or planning for a future deployment:

Example 2. Estimate page cache for the existing Neo4j databases

First, estimate the total size of data and indexes, and then multiply with some factor, for example 20%, to allow for growth.

bin/neo4j-admin server memory-recommendation
...
...
...
# Total size of lucene indexes in all databases: 6690m
# Total size of data and native indexes in all databases: 35050m

You can see that the data volume and native indexes combined take up approximately 35GB. In your specific use case, you estimate that 20% will provide sufficient head room for growth.

server.memory.pagecache.size = 1.2 * (35GB) = 42GB

You configure the page cache by adding the following to neo4j.conf:

server.memory.pagecache.size=42GB

Example 3. Estimate page cache for a new Neo4j database

When planning for a future database, it is useful to run an import with a fraction of the data, and then multiply the resulting store size delta by that fraction plus some percentage for growth.

Run the memory-recommendation command to see the total size of the data and indexes in all current databases.

bin/neo4j-admin server memory-recommendation
...
...
...
# Total size of lucene indexes in all databases: 6690m
# Total size of data and native indexes in all databases: 35050m

Import 1/100th of the data and again measure the data volume and native indexes of all databases.
```
bin/neo4j-admin server memory-recommendation
...
...
...
# Total size of lucene indexes in all databases: 6690m
# Total size of data and native indexes in all databases: 35400m
```
You can see that the data volume and native indexes combined take up approximately 35.4GB.
Multiply the resulting store size delta by that fraction.

35.4GB - 35GB = 0.4GB * 100 = 40GB
Multiply that number by 1.2 to size up the result, and allow for 20% growth.

server.memory.pagecache.size = 1.2 * (40GB) = 48GB
Configure the page cache by adding the following to neo4j.conf:
```
server.memory.pagecache.size=48G
```

Limit transaction memory usage

By using the dbms.memory.transaction.total.max setting you can configure a global maximum memory usage for all of the transactions running on the server. This setting must be configured low enough so that you do not run out of memory. If you are experiencing OutOfMemory messages during high transaction load, try to lower this limit.

Neo4j also offers the following settings to provide fairness, which can help improve stability in multi-tenant deployments.

The setting db.memory.transaction.total.max limits the transaction memory usage per database.
The setting db.memory.transaction.max constrains each transaction.

When any of the limits are reached, the transaction is terminated without affecting the overall health of the database.

To help configure these settings you can use the following commands to list the current usage:

CALL dbms.listPools()
SHOW TRANSACTIONS

Or alternatively, you can monitor the memory usage of each query in the query.log.

Glossary

allocator: A component in the cluster that allocates databases to servers according to the topology constraints specified and an allocation strategy.
asynchronous replication: Asynchronous replication is used by secondary copies to poll for new transactions, which means they cannot be guaranteed to have received the most recent transactions. This enables efficient scale-out of read-performance.
Aura instance: A fully-managed DBMS represented by a single instance ID, that is running in the Neo4j Aura cloud.
auto-commit transaction: An automatically committed transaction that contains a single query.
Bolt protocol: Bolt is a protocol used for interaction between Neo4j instances and drivers.
bookmark: A marker the client can request from the cluster to ensure that it is able to read its own writes so that the application’s state is consistent and only databases that have a copy of the bookmark are permitted to respond.
category (Bloom): A category is based on a node label and is defined in a Perspective as a way of visually distinguishing nodes with the same label(s).
causal consistency: All servers in a cluster agree on the order in which transactions take place. The position of a server on the causal chain can be guaranteed using a bookmark.
cluster: A Neo4j DBMS that spans multiple servers working together to increase fault tolerance and/or read scalability. Databases on a cluster may be configured to replicate across servers in the cluster thus achieving read scalability or high availability.
client application: Software that interacts with a Neo4j server.
commit: A commit is the successful completion of a transaction, which ensures durability of any changes made. For more details, visit Operations Manual → Transaction management.
composite database: Composite databases are the means to access partitioned graph data with a single Cypher query.
constraint: Constraints are sets of data modeling rules that ensure the data is consistent and reliable.
Cypher®: Neo4j’s graph query language.
data model: A data model defines how information is organized in a database. A good data model will make querying and understanding your data easier. In Neo4j, the data models have a graph structure.
database: A database is a container used by the DBMS to manage and store graph data. The physical structure of data is controlled by the database.
database vs graph: Databases are the physical containers of graph data. Graphs are the logical structure of data in Neo4j.
Database Management System: Database Management System, or DBMS, capable of managing multiple databases. A DBMS may run on a single server, or span several servers configured as a cluster.
database schema: The prescribed property existence and datatypes for nodes and relationships.
deallocate: An act of removing a database from a server or a server from a cluster without loss of data or reduced fault tolerance.
degree (of a node): The number of relationships of a specific node; loops are counted twice.
disaster recovery: A manual intervention to restore availability of a cluster, or databases within a cluster.
driver: A software library that provides access to Neo4j from a particular programming language.
election: In the event that the Raft leader becomes unresponsive, followers automatically trigger an election and vote for a new leader.
entity: A node or a relationship.
expression (Cypher): A component of a Cypher query which produces values. It may be used in projections, as a predicate, or when setting properties on graph elements.
fabric: Fabric is the architectural design of a unified system that provides a single access point to local or distributed graph data.
fault tolerance: A guarantee that a cluster can maintain a database’s persistence and availability in the event of one or more servers failing.
follower: A primary copy of a database acting as a follower, receives and acknowledges synchronous writes from the leader.
Generative AI (GenAI): A type of artificial intelligence (AI) system that generates text, images, or other media in response to prompts.
graph: A logical representation of a set of nodes where some pairs are connected by relationships.
index: Data structure that improves read performance of a database.
knowledge graph: A specific type of graph that has an organizing principle so that a user (or a computer system) can reason about the underlying data. The organizing principle provides an additional layer of structure that adds context to support knowledge discovery.
label: Marks a node as a member of a named and indexed subset. A node may be assigned zero or more labels.
leader: A single primary copy of a database is designated as the leader. It receives all write transactions from clients and replicates writes synchronously to followers and asynchronously to secondary copies of the database.
main database: In terms of Neo4j Enterprise Studio, the database(s) containing the user’s data. Can exist in the same Neo4j deployment as the tool asset database.
motif: A description of a specific pattern within a graph.
node: A node represents an entity or discrete object in your graph data model. Nodes can be connected by relationships, hold data in properties, and are classified by labels.
operator: A symbol representing a mathematical or logical operation.
parameter: Named value provided when running a Cypher statement.
path: A sequence of nodes and the relationships connecting them, that does not contain duplicate relationships. Several paths can match a pattern.
pattern: A specific arrangement of nodes and relationships that can be matched in a graph. A pattern follows a motif.
perspective (Bloom): A Perspective defines a certain business view or domain that can be found in the target Neo4j graph. A single Neo4j graph can be viewed through different Perspectives, each tailored for a different business purpose.
primary: A copy of the database that is able to process write transactions and is eligible to be elected as a leader. It participates in fault tolerant writes as it is part of the majority required to acknowledge and commit write transactions.
primary vs secondary: In a cluster, databases can operate in either primary or secondary mode. Primary databases are able to process write and read transactions, ensuring fault tolerance. Secondary databases are replicated asynchronously from primaries, and their main purpose is to provide read scaling within the cluster.
project (Aura): An isolated environment in the unified Aura console that contains its own database instances, configurations, and resources. Preceded by tenant in the classic Aura console.
property: Properties are key-value pairs that are used for storing data on nodes and relationships.
query (Cypher): A statement that retrieves or writes information to a database.
Raft group: A group of servers that are participating in hosting a particular database in primary mode.
Raft group member: A server that is participating in a Raft group. A server can be a member of one or more groups.
Raft log: A shared log between all Raft group members that is guaranteed to be consistently updated and viewed by those members. The log contains both database data and operational state of the Raft group.
Raft protocol: The networking mechanism that enables a database to replicate its data across multiple servers to give high availability for accessing the data and high durability to the data stored.
read scaling: Distributing query load by creating additional database copies hosted in secondary mode (read-only).
relationship: A relationship represents a connection between nodes in your graph data model. Relationships connect a source node to a target node, hold data in properties, and are classified by type.
secondary: An asynchronously replicated copy of the database that provides read scaling within the cluster.
seed: A seed is a database dump or a full backup used to create a database on a cluster. This is sometimes called seeding.
server: A physical machine, a virtual machine, or a container running an instance of Neo4j. Servers can be standalone or part of a cluster.
session: A causally linked sequence of transactions.
session consistency: An alternative name for Neo4j’s causal consistency.
standalone: A single server running Neo4j and not part of a cluster.
synchronous replication: Synchronous replication requires the leader primary to replicate a transaction and block the commit until a quorum of the follower primaries acknowledges that the transaction is successfully replicated. Once the transaction is replicated, the commit is allowed to proceed. This ensures data durability and consistency within the cluster.
system database: A database used by Neo4j to store system information.
tenant (Aura): An isolated environment in the classic Aura console that contains its own database instances, configurations, and resources. Replaced by project in the unified Aura console.
tool asset database: In terms of Neo4j Enterprise Studio, the database where tools' assets are stored. This can be in the same Neo4j deployment as the main database(s) or in a separate deployment.
topology: A configuration that describes how the copies of a database should be spread across the servers in a cluster, see primary mode and secondary mode.
transaction: A transaction comprises a unit of work performed against a database. It is treated in a coherent and reliable way, independent of other transactions. Transactions comply with the ACID consistency model (atomic, consistent, isolated, and durable).