Performance Considerations
On this page
Overview
In this guide, you can learn how to optimize the performance of the
Rust driver. To connect to MongoDB, you must create a Client
instance. Your Client
instance automatically handles most aspects of connection, such as
discovering server topology, monitoring your connection, and maintaining
an internal connection pool. This guide describes best practices for
configuring and using your Client
instance.
This guide includes the following sections:
Client Lifecycle describes best practices for creating and managing a
Client
instanceConnection Pool describes how connection pooling works in the driver
Parallelism provides sample code for running parallel, asynchronous tasks
Runtime describes how to manage runtimes by using functionalities of the
tokio
andasync_std
cratesAdditional Information provides links to resources and API documentation for types and methods mentioned in this guide
Client Lifecycle
We recommend that you reuse your client across sessions and operations.
You can use the same Client
instance to perform multiple tasks, as
the Client
type is safe for concurrent use by multiple threads.
Creating a new Client
instance for each request results in slower
performance.
The following code creates a method that accepts a pointer to an
existing Client
instance, which allows you to perform many requests
by using the same client:
// ... Create a client earlier in your code async fn make_request(client: &Client) -> Result<(), Box<dyn Error>> { // Use the client to perform operations Ok(()) }
Connection Pool
Every Client
instance has a built-in connection pool for each server
in your MongoDB topology. Connection pools open sockets on demand to
support concurrent requests to MongoDB in your application.
The default configuration for a Client
works for most applications.
The following code shows how to create a client with default connection
settings:
let client = Client::with_uri_str("<connection string>").await?;
Alternatively, you can tune the connection pool to best fit the needs of your application and optimize performance. For more information on how to customize your connection settings, see the following subsections of this guide:
Tip
To learn more about configuring a connection pool, see Tuning Your Connection Pool Settings in the Server manual.
Configure Maximum Pool Size
The maximum size of each connection pool is set by the max_pool_size
option, which defaults to 10
. If the number of in-use connections to
a server reaches the value of max_pool_size
, the next request to
that server waits until a connection becomes available.
In addition to the sockets needed to support your application's requests,
each Client
instance opens two more sockets per server
in your MongoDB topology for monitoring the server's state.
For example, a client connected to a three-node replica set opens six
monitoring sockets. If the application uses the default setting for
max_pool_size
and only queries the primary (default) node, then
there can be at most 16 total connections in the connection pool. If the
application uses a read preference to query the secondary nodes, those
connection pools grow and there can be 36 total connections.
To support high numbers of concurrent MongoDB requests
within one process, you can increase the value of the max_pool_size
option. The following code demonstrates how to specify a value for
max_pool_size
when instantiating a Client
:
let mut client_options = ClientOptions::parse("<connection string>").await?; client_options.max_pool_size = Some(20); let client = Client::with_options(client_options)?;
Configure Concurrent Connection Options
Connection pools are rate-limited. The max_connecting
option
determines the number of connections that the pool can create in
parallel at any time. For example, if the value of max_connecting
is
2
, the default value, the third request that attempts to concurrently
check out a connection succeeds only when one of the following cases occurs:
The connection pool finishes creating a connection and the number of connections in the pool is less than or equal to the value of
max_pool_size
.An existing connection is checked back into the pool.
The driver's ability to reuse existing connections improves due to rate limits on connection creation.
You can set the minimum number of concurrent connections to
each server with the min_pool_size
option, which defaults to 0
.
The driver initializes the connection pool with this number of sockets. If
sockets are closed and the total number of sockets, both in use and
idle, drops below the minimum, the connection pool opens more sockets until the
minimum is reached.
The following code sets the max_connecting
and min_pool_size
options when
instantiating a Client
:
let mut client_options = ClientOptions::parse("<connection string>").await?; client_options.max_connecting = Some(3); client_options.min_pool_size = Some(1); let client = Client::with_options(client_options)?;
Configure Maximum Idle Time
You can set the maximum amount of time that a connection can
remain idle in the pool by setting the max_idle_time
option.
Once a connection has been idle for the duration specified in
max_idle_time
, the connection pool removes and replaces that
connection. This option defaults to 0
, or no limit.
When the Client::shutdown()
method is called at any point in your
application, the driver closes all idle sockets and closes all sockets
that are in use as they are returned to the pool. Calling Client::shutdown()
closes only inactive sockets, so you cannot interrupt or terminate
any ongoing operations by using this method. The driver closes these
sockets only when the process completes.
The following code sets the value of the max_idle_time
option to
90
seconds when instantiating a Client
:
let mut client_options = ClientOptions::parse("<connection string>").await?; client_options.max_idle_time = Some(Duration::new(90, 0)); let client = Client::with_options(client_options)?;
Parallelism
If you can run parallel data operations, you can optimize performance by
running asynchronous, concurrent tasks. The following code uses the
spawn()
method from the tokio::task
module to create separate,
concurrent tasks to perform insert operations:
let client = Client::with_uri_str("<connection string>").await?; let data = doc! { "title": "1984", "author": "George Orwell" }; for i in 0..5 { let client_ref = client.clone(); let data_ref = data.clone(); task::spawn(async move { let collection = client_ref .database("items") .collection::<Document>(&format!("coll{}", i)); collection.insert_one(data_ref).await }); }
Runtime
A Client
instance is bound to the instance of the tokio
or
async-std
runtime in which you created it. If you use a Client
instance to perform operations on a different runtime, you might
experience unexpected behavior or failures.
If use the test
helper macro from the tokio
or
async_std
crate to test your application, you might accidentally run
operations on a different runtime than intended. This is because these
helper macros create a new runtime for each test. However, you can use
one of the following strategies to avoid this issue:
Attach the runtime to the
Client
instance without using thetest
helper macros.Create a new
Client
instance for everyasync
test.
This example follows the first strategy and creates a global runtime used only for testing.
In the following code, the test_list_dbs()
method uses a client that
manually connects to this runtime to list databases in the deployment:
use tokio::runtime::Runtime; use once_cell::sync::Lazy; static CLIENT_RUNTIME: Lazy<(Client, Runtime)> = Lazy::new(|| { let rt = Runtime::new().unwrap(); let client = rt.block_on(async { Client::with_uri_str("<connection string>").await.unwrap() }); (client, rt) }); fn test_list_dbs() -> Result<(), Box<dyn Error>> { let (client, rt) = &*CLIENT_RUNTIME; rt.block_on(async { client.list_database_names().await })?; Ok(()) }
Implementing the second strategy, the following code creates a new
Client
instance for each test run with tokio::test
,
ensuring that there are no unintended interactions between runtimes:
async fn test_list_dbs() -> Result<(), Box<dyn Error>> { let client = Client::with_uri_str("<connection string>").await?; client.list_database_names().await?; Ok(()) }
Additional Information
To learn more about connecting to MongoDB, see the Connection Guide.
To learn more about the available runtimes for the Rust driver, see the guide on Asynchronous and Synchronous APIs.
API Documentation
spawn() in the
tokio::task
moduletokio::runtime module