Step 6: (Optional) Best practices to configure the connection pool size for your application - Amazon Keyspaces (for Apache Cassandra)
Services or capabilities described in Amazon Web Services documentation might vary by Region. To see the differences applicable to the China Regions, see Getting Started with Amazon Web Services in China (PDF).

Step 6: (Optional) Best practices to configure the connection pool size for your application

In this section, we outline how to determine the ideal connection pool size based on the query throughput requirements of your application.

Amazon Keyspaces allows a maximum of 3,000 CQL queries per second per TCP connection. So there's virtually no limit on the number of connections that a driver can establish with Amazon Keyspaces. However, we recommend that you match the connection pool size to the requirements of your application and consider the available endpoints when you're using Amazon Keyspaces with VPC endpoint connections.

You configure the connection pool size in the client driver. For example, based on a local pool size of 2 and a VPC interface endpoint created across 3 Availability Zones, the driver establishes 6 connections for querying (7 in total, which includes a control connection). Using these 6 connections, you can support a maximum of 18,000 CQL queries per second.

If your application needs to support 40,000 CQL queries per second, work backwards from the number of queries that are needed to determine the required connection pool size. To support 40,000 CQL queries per second, you need to configure the local pool size to be at least 5, which supports a minimum of 45,000 CQL queries per second.

You can monitor if you exceed the quota for the maximum number of operations per second, per connection by using the PerConnectionRequestRateExceeded CloudWatch metric in the Amazon/Cassandra namespace. The PerConnectionRequestRateExceeded metric shows the number of requests to Amazon Keyspaces that exceed the quota for the per-connection request rate.

The code examples in this step show how to estimate and configure connection pooling when you're using interface VPC endpoints.

Java

You can configure the number of connections per pool in the Java driver. For a complete example of a Java client driver connection, see Using a Cassandra Java client driver to access Amazon Keyspaces programmatically.

When the client driver is started, first the control connection is established for administrative tasks, such as for schema and topology changes. Then the additional connections are created.

In the following example, the local pool size driver configuration is specified as 2. If the VPC endpoint is created across 3 subnets within the VPC, this results in 7 NewConnections in CloudWatch for the interface endpoint, as shown in the following formula. 

NewConnections = 3 (VPC subnet endpoints created across) * 2 (pool size) + 1 ( control connection)
datastax-java-driver {

    basic.contact-points = [ "cassandra.us-east-1.amazonaws.com:9142"]
    advanced.auth-provider{
        class = PlainTextAuthProvider
           username = "ServiceUserName"
           password = "ServicePassword"
    }
    basic.load-balancing-policy {
        local-datacenter = "us-east-1"
        slow-replica-avoidance = false           
    }

    advanced.ssl-engine-factory {
        class = DefaultSslEngineFactory
        truststore-path = "./src/main/resources/cassandra_truststore.jks"
        truststore-password = "my_password"
        hostname-validation = false
      }
    advanced.connection {
        pool.local.size = 2
          }
}

If the number of active connections doesn’t match your configured pool size (aggregation across subnets) + 1 control connection, something is preventing the connections from being created.

Node.js

You can configure the number of connections per pool in the Node.js driver. For a complete example of a Node.js client driver connection, see Using a Cassandra Node.js client driver to access Amazon Keyspaces programmatically.

For the following code example, the local pool size driver configuration is specified as 1. If the VPC endpoint is created across 4 subnets within the VPC, this results in 5 NewConnections in CloudWatch for the interface endpoint, as shown in the following formula.

NewConnections = 4 (VPC subnet endpoints created across) * 1 (pool size) + 1 ( control connection)
const cassandra = require('cassandra-driver');
const fs = require('fs');
const types = cassandra.types;
const auth = new cassandra.auth.PlainTextAuthProvider('ServiceUserName', 'ServicePassword');
const sslOptions1 = {
         ca: [
                    fs.readFileSync('/home/ec2-user/sf-class2-root.crt', 'utf-8')],
                    host: 'cassandra.us-east-1.amazonaws.com',
                    rejectUnauthorized: true
        };
const client = new cassandra.Client({
                   contactPoints: ['cassandra.us-east-1.amazonaws.com'],
                   localDataCenter: 'us-east-1',
                   pooling: { coreConnectionsPerHost: { [types.distance.local]: 1 } },
                   consistency: types.consistencies.localQuorum,
                   queryOptions: { isIdempotent: true },
                   authProvider: auth,
                   sslOptions: sslOptions1,
                   protocolOptions: { port: 9142 }
        });