Kafka Cluster Types in Confluent Cloud

Confluent offers different types of Kafka clusters in Confluent Cloud. The cluster type you choose determines the features, capabilities, and price of the cluster. Use the information in this topic to find the cluster with the features and capabilities that best meets your needs. Use Basic clusters for experimentation and early development. For a production cluster, choose from Standard, Enterprise, or Dedicated.

The table below offers a high-level comparison of features across Kafka cluster types.

Feature	Basic	Standard	Enterprise	Dedicated
Kafka ACLs	Yes	Yes	Yes	Yes
Encryption-at-rest	Yes	Yes	Yes	Yes
Encryption-in-transit	Yes	Yes	Yes	Yes
Exactly Once Semantics	Yes	Yes	Yes	Yes
Fully-managed replica placement	Yes	Yes	Yes	Yes
Key based compacted storage	Yes	Yes	Yes	Yes
User interface to manage consumer lag	Yes	Yes	Yes	Yes
Topic management	Yes	Yes	Yes	Yes
Uptime Service Level Agreement (SLA)	Yes	Yes	Yes	Yes
Role-based Access Control (RBAC)	Yes [*]	Yes	Yes	Yes
OAuth	No	Yes	Yes	Yes
Audit logs	No	Yes	Yes	Yes
Public networking	Yes	Yes	No	Yes
Private networking	No	No	Yes	Yes
Self-managed encryption keys	No	No	No	Yes

[*]	RBAC roles for resources within the Kafka cluster (DeveloperRead, DeveloperWrite, DeveloperManage, and ResourceOwner) are not available on Basic clusters.

Important

The capabilities provided in this topic are for planning purposes, and are not a guarantee of performance, which varies depending on each unique configuration.

Basic clusters

Basic Kafka clusters are designed for experimentation, early development, and basic use cases. Basic clusters support the following:

• 99.5% uptime SLA
• Up to 1 GB/s of throughput and 5 TB of storage.
• You only pay for the ingress, egress, storage, and partitions. There is no base cluster price.
• Can be upgraded to a single-zone Standard cluster at any time using the Confluent Cloud Console.

Note

You can view connector events in Confluent Cloud Console with a Basic cluster, but you can’t consume events from a topic using Confluent CLI, Java, or C/C++. For more information, see View Connector Events.

Basic limits per cluster

Dimension	Capability	Additional details
Ingress	Max 250 MBps	Number of bytes that can be produced to the cluster in one second. Available in the Metrics API as received_bytes (convert from bytes to MB). If you are self-managing Kafka, you can look at the producer outgoing-byte-rate metrics and broker kafka.server:type=BrokerTopicMetrics,name=BytesInPerSec metrics to understand your throughput. To reduce usage on this dimension, you can compress your messages. lz4 is recommended for compression. gzip is not recommended because it incurs high overhead on the cluster.
Egress	Max 750 MBps	Number of bytes that can be consumed from the cluster in one second. Available in the the Metrics API as sent_bytes (convert from bytes to MB). If you are self-managing Kafka, you can look at the consumer incoming-byte-rate metrics and broker kafka.server:type=BrokerTopicMetrics,name=BytesOutPerSec to understand your throughput. To reduce usage on this dimension, you can compress your messages and ensure each consumer is only consuming from the topics it requires. lz4 is recommended for compression. gzip is not recommended because it incurs high overhead on the cluster.
Storage (pre-replication)	Max 5 TB	Number of bytes retained on the cluster, pre-replication. Available in the Metrics API as retained_bytes (convert from bytes to TB). The returned value is pre-replication. You can configure policy settings retention.bytes and retention.ms at a topic level so you can control exactly how much and how long to retain data in a way that makes sense for your applications and helps control your costs. If you are self-managing Kafka, you can look at how much disk space your cluster is using to understand your storage needs. To reduce usage on this dimension, you can compress your messages and reduce your retention settings. lz4 is recommended for compression. gzip is not recommended because it incurs high overhead on the cluster.
Partitions (pre-replication)	Max 4096	Maximum number of partitions that can exist on the cluster at one time, before replication. All topics that the customer creates as well as internal topics that are automatically created by Confluent Platform components–such as ksqlDB, Kafka Streams, Connect, and Control Center–count towards the cluster partition limit. The automatically created topics are prefixed with an underscore (_). Topics that are internal to Kafka itself (e.g., consumer offsets) are not visible in the Cloud Console, and do not count against partition limits or toward partition billing. Available in the Metrics API as partition_count. Attempts to create additional partitions beyond this limit will fail with an error message. If you are self-managing Kafka, you can look at the kafka.controller:type=KafkaController,name=GlobalPartitionCount metric to understand your partition usage. Find details in the Broker section. To reduce usage on this dimension, you can delete unused topics and create new topics with fewer partitions. You can use the Kafka Admin interface to increase the partition count of an existing topic if the initial partition count is too low.
Total client connections	Max 1000	Number of TCP connections to the cluster that can be open at one time. Available in the Metrics API as active_connection_count. Filter by principal to understand how many connections each application is creating. If you are self-managing Kafka, you can look at the broker kafka.server:type=socket-server-metrics,listener={listener_name},networkProcessor={#},name=connection-count metrics to understand how many connections you are using. This value may not have a 1:1 ratio to connections in Confluent Cloud, depending on the number of brokers, partitions, and applications in your self-managed cluster. How many connections a cluster supports can vary widely based on several factors, including number of producer clients, number of consumer clients, partition keying strategy, produce patterns per client, and consume patterns per client.
Connection attempts	Max 80 per second	Maximum number of new TCP connections to the cluster that can be created in one second. This means successful authentications plus unsuccessful authentication attempts. Available in the Metrics API as successful_authentication_count (only includes successful authentications, not unsuccessful authentication attempts). If you are self-managing Kafka, you can look at the rate of change for the kafka.server:type=socket-server-metrics,listener={listener_name},networkProcessor={#},name=connection-count metric and the Consumer connection-creation-rate metric to understand how many new connections you are creating. For details, see Broker Metrics and Global Connection Metrics. To reduce usage on this dimension, you can use longer-lived connections to the cluster.
Requests	Max ~15000 per second	Number of client requests to the cluster in one second. Available in the the Metrics API as request_count. If you are self-managing Kafka, you can look at the broker kafka.network:type=RequestMetrics,name=RequestsPerSec,request={Produce FetchConsumer FetchFollower} metrics and client request-rate metrics to understand your request volume. To reduce usage on this dimension, you can adjust producer batching configurations, consumer client batching configurations, and shut down otherwise inactive clients.
Message size	Max 8 MB	For message size defaults at the topic level see Kafka topic configurations for all Confluent Cloud cluster types.
Client version	Minimum 0.11.0	None
Request size	Max 100 MB	None
Fetch bytes	Max 55 MB	None
API keys	Default limit 50	You can request an increase. For more information, see Service Quotas for Confluent Cloud.
Partition creation and deletion	Max 250 per 5 minute period	The following occurs when partition creation and deletion rate limit is reached: For clients < Kafka 2.7: The cluster always accepts and processes all the partition creates and deletes within a request, and then throttles the connection of the client until the rate of changes is below the quota. For clients >= Kafka 2.7: The cluster only accepts and processes partition creates and deletes up to the quota. All other partition creates and deletes in the request are rejected with a THROTTLING_QUOTA_EXCEEDED error. By default, the admin client will automatically retry on that error until default.api.timeout.ms is reached. When the automatic retry is disabled by the client, the THROTTLING_QUOTA_EXCEEDED error is immediately returned to the client.
Connector tasks per Kafka cluster	Max 250	1 task per connector
ACLs	Default limit 1000	You can request an increase. For more information, see Service Quotas for Confluent Cloud.
Kafka REST Produce v3	Max throughput: 10 MBps Max connection requests: 25 connection requests per second Max streamed requests: 1000 requests per second Max message size for Kafka REST Produce API: 1 MB	Client applications can connect over the REST API to Produce records directly to the Confluent Cloud cluster. To learn more, see the Kafka REST concepts overview. To learn more about the Kafka REST Produce API streaming mode, see the examples and explanation of streaming mode in the concept docs, and the Produce example in the quick start. A streamed request is a single record to be produced to Kafka. Concatenate the records to produce multiple records in the same request. Delivery reports are concatenated in the same order as the records are sent. For more information, see Produce Records.

Basic limits per partition

The partition capabilities that follow are based on benchmarking and intended as practical guidelines for planning purposes. Performance per partition will vary depending on your individual configuration, and these benchmarks do not guarantee performance.

Dimension	Capability
Ingress per Partition	~5 MBps
Egress per Partition	~15 MBps
Storage per Partition	Unlimited
Storage per Partition for Compacted Topics	Unlimited

Cluster Linking capabilities

• A Basic cluster can be a source cluster of a cluster link (can send data).
• A Basic cluster cannot be a destination cluster (cannot receive data).

To learn more, see Supported cluster types in the Cluster Linking documentation.

Standard clusters

Standard clusters are designed for production-ready features and functionality. Standard clusters support the following:

• Uptime SLA: 99.95% for Single-Zone, 99.99% for Multi-Zone
• Up to 1 GB/s of throughput and unlimited storage.
• Multi-zone high availability (optional). A multi-zone cluster is spread across three availability zones for better resiliency.
• Charged an hourly base price in addition to the ingress, egress, storage, and partitions.

Standard limits per cluster

Dimension	Capability	Additional details
Ingress	Max 250 MBps	Number of bytes that can be produced to the cluster in one second. Available in the Metrics API as received_bytes (convert from bytes to MB). If you are self-managing Kafka, you can look at the producer outgoing-byte-rate metrics and broker kafka.server:type=BrokerTopicMetrics,name=BytesInPerSec metrics to understand your throughput. To reduce usage on this dimension, you can compress your messages. lz4 is recommended for compression. gzip is not recommended because it incurs high overhead on the cluster.
Egress	Max 750 MBps	Number of bytes that can be consumed from the cluster in one second. Available in the the Metrics API as sent_bytes (convert from bytes to MB). If you are self-managing Kafka, you can look at the consumer incoming-byte-rate metrics and broker kafka.server:type=BrokerTopicMetrics,name=BytesOutPerSec to understand your throughput. To reduce usage on this dimension, you can compress your messages and ensure each consumer is only consuming from the topics it requires. lz4 is recommended for compression. gzip is not recommended because it incurs high overhead on the cluster.
Storage (pre-replication)	Infinite	Number of bytes retained on the cluster, pre-replication. Standard Confluent Cloud clusters support infinite storage. This means there is no maximum size limit for the amount of data that can be stored on the cluster. Available in the Metrics API as retained_bytes (convert from bytes to TB). The value returned is pre-replication. You can configure policy settings retention.bytes and retention.ms at a topic level so you can control exactly how much and how long to retain data in a way that makes sense for your applications and helps control your costs. If you exceed configured maximum retention values, producers will be throttled to prevent additional writes, which will register as non-zero values for the producer client produce-throttle-time-max and produce-throttle-time-avg metrics. If you are self-managing Kafka, you can look at how much disk space your cluster is using to understand your storage needs. To reduce usage on this dimension, you can compress your messages and reduce your retention settings. lz4 is recommended for compression. gzip is not recommended because it incurs high overhead on the cluster.
Partitions (pre-replication)	Max 4096	Maximum number of partitions that can exist on the cluster at one time, before replication. All topics that the customer creates as well as internal topics that are automatically created by Confluent Platform components–such as ksqlDB, Kafka Streams, Connect, and Control Center–count towards the cluster partition limit. The automatically created topics are prefixed with an underscore (_). Topics that are internal to Kafka itself (e.g., consumer offsets) are not visible in the Cloud Console, and do not count against partition limits or toward partition billing. Available in the Metrics API as partition_count. Attempts to create additional partitions beyond this limit will fail with an error message. If you are self-managing Kafka, you can look at the kafka.controller:type=KafkaController,name=GlobalPartitionCount metric to understand your partition usage. Find details in the Broker section. To reduce usage on this dimension, you can delete unused topics and create new topics with fewer partitions. You can use the Kafka Admin interface to increase the partition count of an existing topic if the initial partition count is too low.
Total client connections	Max 1000	Number of TCP connections to the cluster that can be open at one time. Available in the Metrics API as active_connection_count. Filter by principal to understand how many connections each application is creating. If you are self-managing Kafka, you can look at the broker kafka.server:type=socket-server-metrics,listener={listener_name},networkProcessor={#},name=connection-count metrics to understand how many connections you are using. This value may not have a 1:1 ratio to connections in Confluent Cloud, depending on the number of brokers, partitions, and applications in your self-managed cluster. How many connections a cluster supports can vary widely based on several factors, including number of producer clients, number of consumer clients, partition keying strategy, produce patterns per client, and consume patterns per client.
Connection attempts	Max 80 per second	Maximum number of new TCP connections to the cluster that can be created in one second. This means successful authentications plus unsuccessful authentication attempts. Available in the Metrics API as successful_authentication_count (only includes successful authentications, not unsuccessful authentication attempts). If you are self-managing Kafka, you can look at the rate of change for the kafka.server:type=socket-server-metrics,listener={listener_name},networkProcessor={#},name=connection-count metric and the Consumer connection-creation-rate metric to understand how many new connections you are creating. For details, see Broker Metrics and Global Connection Metrics. To reduce usage on this dimension, you can use longer-lived connections to the cluster.
Requests	Max ~15000 per second	Number of client requests to the cluster in one second. Available in the the Metrics API as request_count. If you are self-managing Kafka, you can look at the broker kafka.network:type=RequestMetrics,name=RequestsPerSec,request={Produce FetchConsumer FetchFollower} metrics and client request-rate metrics to understand your request volume. To reduce usage on this dimension, you can adjust producer batching configurations, consumer client batching configurations, and shut down otherwise inactive clients.
Message size	Max 8 MB	For message size defaults at the topic level, see Kafka topic configurations for all Confluent Cloud cluster types.
Client version	Minimum 0.11.0	None
Request size	Max 100 MB	None
Fetch bytes	Max 55 MB	None
API keys	Default limit 100	You can request an increase. For more information, see Service Quotas for Confluent Cloud.
Partition creation and deletion	Max 500 per 5 minute period	The following occurs when partition creation and deletion rate limit is reached: For clients < Kafka 2.7: The cluster always accepts and processes all the partition creates and deletes within a request, and then throttles the connection of the client until the rate of changes is below the quota. For clients >= Kafka 2.7: The cluster only accepts and processes partition creates and deletes up to the quota. All other partition creates and deletes in the request are rejected with a THROTTLING_QUOTA_EXCEEDED error. By default, the admin client will automatically retry on that error until default.api.timeout.ms is reached. When the automatic retry is disabled by the client, the THROTTLING_QUOTA_EXCEEDED error is immediately returned to the client.
Connector tasks per Kafka cluster	Max 250	None
ACLs	Default limit 1000	You can request an increase. For more information, see Service Quotas for Confluent Cloud.
Kafka REST Produce v3	Max throughput: 10 MBps Max connection requests: 25 connection requests per second Max streamed requests: 1000 requests per second Max message size for Kafka REST Produce API: 1 MB	Client applications can connect over the REST API to Produce records directly to the Confluent Cloud cluster. To learn more, see the Kafka REST concepts overview. To learn more about the Kafka REST Produce API streaming mode, see the examples and explanation of streaming mode in the concept docs, and the Produce example in the quick start. A streamed request is a single record to be produced to Kafka. Concatenate the records to produce multiple records in the same request. Delivery reports are concatenated in the same order as the records are sent. For more information, see Produce Records.

Standard limits per partition

The partition capabilities that follow are based on benchmarking and intended as practical guidelines for planning purposes. Performance per partition will vary depending on your individual configuration, and these benchmarks do not guarantee performance.

Dimension	Capability
Ingress per Partition	~5 MBps
Egress per Partition	~15 MBps
Storage per Partition	Unlimited
Storage per Partition for Compacted Topics	Unlimited

Cluster Linking capabilities

• A Standard cluster can be a source cluster of a cluster link (can send data).
• A Standard cluster cannot be a destination cluster (cannot receive data).

To learn more, see Supported cluster types in the Cluster Linking documentation.

Enterprise clusters

Enterprise clusters are designed for production-ready functionality that requires private endpoint networking capabilities. Enterprise clusters support the following:

• Uptime SLA: 99.99%
• Up to 1 GB/s of throughput and unlimited storage
• Multi-zone high availability cluster spread across three availability zones for better resiliency
• Automatic scaling based on load

Confluent uses Confluent Unit for Kafka for Enterprise clusters (E-CKU) to provision and bill for Enterprise Kafka clusters. Sometimes E-CKU is referred to as CKU for Enterprise.

E-CKU limits per cluster

Enterprise clusters are elastic, shrinking and expanding automatically based on load. You don’t resize your cluster. When you need more capacity, your Enterprise cluster expands up to the fixed ceiling. If you’re not using capacity above the minimum, you’re not paying for it.

Enterprise cluster capacity:

• Minimum: two E-CKU
• Fixed ceiling: five E-CKU

E-CKU capacity guidance

The dimensions in the following table describe the capacity of a single E-CKU. For more information about E-CKU, see Confluent Unit for Kafka for Enterprise clusters and Compare CKU for Enterprise and Dedicated Kafka clusters.

Dimension	E-CKU capacity
Storage (pre-replication)	Infinite
Partitions (pre-replication)	1000 partitions for non-compacted topics, 120 for compacted topics
Connection attempts	250 connection attempts per second
Ingress	~50 megabytes per second (MBps)
Egress	~150 megabytes per second (MBps)
Total client connections	~4,500 connections
Requests	~7,500 requests per second

Enterprise limits per cluster

Enterprise clusters have a maximum capacity of five E-CKUs. For any Confluent Cloud cluster, the expected performance for any given workload is dependent on a variety of dimensions, such as message size and number of partitions.

Use the information in the following tables to determine if a given workload fits within the fixed ceiling, how to monitor dimensions, and suggestions to reduce your use of a particular dimension.

Dimension	Limits per cluster
Ingress	250 megabytes per second (MBps)
Egress	750 megabytes per second (MBps)
Total client connections	22,500 connections
Requests	37,500 requests per second
Storage (pre-replication)	Infinite
Partitions (pre-replication)	5,000 partitions for non-compacted topics, 600 for compacted topics
Connection attempts	1,250 connection attempts per second
Kafka REST Produce v3	Max throughput: 10 MBps Max connection requests: 25 connection requests per second Max streamed requests: 1000 requests per second Max message size for Kafka REST Produce API: 1 MB
Message size	Max 8 MB
Client version	Minimum 0.11.0
Request size	Max 100 MB
Fetch bytes	Max 55 MB
API keys	Default limit 500 (You can request an increase. For more information, see Service Quotas for Confluent Cloud.)
Partition creation and deletion	Max 5000 per 5 minute period
Connector tasks per Kafka cluster	Max 250
ACLs	Default limit 4,000 (You can request an increase. For more information, see Service Quotas for Confluent Cloud.)

Enterprise limits per partition

The partition capabilities that follow are based on benchmarking and intended as practical guidelines for planning purposes. Performance per partition will vary depending on your individual configuration, and these benchmarks do not guarantee performance.

Dimension	Capability
Ingress per Partition	~5 MBps
Egress per Partition	~15 MBps
Storage per Partition	Unlimited
Storage per Partition for Compacted Topics	Unlimited

Cluster Linking capabilities

Enterprise clusters can be a source of a cluster link, dependent on the networking type and the other cluster involved. To learn more, see Supported cluster types in the Cluster Linking documentation.

Dedicated clusters

Dedicated clusters are designed for critical production workloads with high traffic or private networking requirements. Dedicated clusters support the following:

• Single-tenant deployments with a 99.95% uptime SLA for Single-Zone, and 99.99% for Multi-Zone
• Private networking options including VPC peering, AWS Transit Gateway, AWS PrivateLink, and Azure PrivateLink.
• Self-managed keys when AWS, Azure, or Google Cloud is the cloud service provider.
• Multi-zone high availability (optional). A multi-zone cluster is spread across three availability zones for better resiliency.
• Can be scaled to achieve gigabytes per second of ingress.
• Simple scaling in terms of CKUs.
• Cluster expansion, and Cluster shrinking.
• Cluster Linking for fully-managed replication (multiregion, multicloud, hybrid cloud, and inter-organization).

Dedicated clusters are provisioned and billed in terms of Confluent Unit for Kafka (CKU). CKUs are a unit of horizontal scalability in Confluent Cloud that provide a pre-allocated amount of resources. How much you can ingest and stream per CKU depends on a variety of factors including client application design and partitioning strategy.

CKU limits per cluster

Dedicated clusters can be purchased in any whole number of CKUs up to a limit.

• For organizations with credit card billing, the upper limit is 4 CKUs per Dedicated cluster. Clusters up to 100 CKUs are available by request.
• For organizations with integrated cloud provider billing or payment using an invoice, the upper limit is 24 CKUs per Dedicated cluster. Clusters up to 100 CKUs are available by request.

For clusters that can scale to 100 CKU, contact Confluent Support to discuss the onboarding process and product considerations.

Single-zone clusters can have 1 or more CKUs, whereas multi-zone clusters, which are spread across three availability zones, require a minimum of 2 CKUs. Zone availability cannot be changed after the cluster is created.

Limits per CKU

CKUs determine the capacity of your cluster. For a Confluent Cloud cluster, the expected performance for any given workload is dependent on a variety of dimensions, such as message size and number of partitions.

There are two categories of CKU dimensions:

• Dimensions with a fixed limit that cannot be exceeded.
• Dimensions with a more flexible guideline that may be exceeded depending on the overall cluster load.

The recommended guideline for a dimension is calculated for a workload optimized across the dimensions, enabling high levels of CKU utilization as measured by the cluster load metric. You may exceed the recommended guideline for a dimension, and achieve higher performance for that dimension, usually only if your usage of other dimensions is less than the recommended guideline or fixed limit.

Also note that usage patterns across all dimensions affect the workload and you may not achieve the suggested guideline for a particular dimension. For example, if you reach the partition limit, you will not likely reach the maximum CKU throughput guideline.

You should monitor the cluster load metric for your cluster to see how your usage pattern correlates with cluster utilization.

When a cluster’s load metric is high, the cluster may delay new connections and/or throttle clients in an attempt to ensure the cluster remains available. This throttling would register as non-zero values for the producer client produce-throttle-time-max and produce-throttle-time-avg metrics and consumer client fetch-throttle-time-max and fetch-throttle-time-avg metrics.

Use the information in the following tables to determine the minimum number of CKUs to use for a given workload, how to monitor a dimension and suggestions to reduce your use of a particular dimension.

Dimensions with fixed limits

The following table lists dimensions that have a fixed maximum limit that cannot be exceeded.

Dimension	Maximum per CKU	Details
Storage (pre-replication)	Infinite	Number of bytes retained on the cluster, pre-replication. Dedicated clusters have infinite storage, which means there is no maximum size limit for the amount of data that can be stored on the cluster. Available in the Metrics API as retained_bytes (convert from bytes to TB). The returned value is pre-replication. You can configure policy settings retention.bytes and retention.ms at a topic level so you can control exactly how much and how long to retain data in a way that makes sense for your applications and helps control your costs. If configured retention values are exceeded, producers will be throttled to prevent additional writes, which will register as non-zero values for the producer client produce-throttle-time-max and produce-throttle-time-avg metrics. If you are self-managing Kafka, you can look at how much disk space your cluster is using to understand your storage needs. To reduce usage on this dimension, you can compress your messages and reduce your retention settings. lz4 is recommended for compression. gzip is not recommended because it incurs high overhead on the cluster.
Partitions (pre-replication)	4,500 partitions (100,000 partitions maximum across all CKUs)	Maximum number of partitions that can exist on the cluster at one time, before replication. All topics that the customer creates as well as internal topics that are automatically created by Confluent Platform components–such as ksqlDB, Kafka Streams, Connect, and Control Center–count towards the cluster partition limit. The automatically created topics are prefixed with an underscore (_). Topics that are internal to Kafka itself (e.g., consumer offsets) are not visible in the Cloud Console, and do not count against partition limits or toward partition billing. Available in the Metrics API as partition_count. Attempts to create additional partitions beyond this limit will fail with an error message. If you are self-managing Kafka, you can look at the kafka.controller:type=KafkaController,name=GlobalPartitionCount metric to understand your partition usage. Find details in the Broker section. To reduce usage on this dimension, you can delete unused topics and create new topics with fewer partitions. You can use the Kafka Admin interface to increase the partition count of an existing topic if the initial partition count is too low.
Connection attempts	500 connection attempts per second	Maximum number of new TCP connections to the cluster that can be created in one second. This means successful authentications plus unsuccessful authentication attempts. If you exceed the maximum, connection attempts may be refused. If you are self-managing Kafka, you can look at the rate of change for the kafka.server:type=socket-server-metrics,listener={listener_name},networkProcessor={#},name=connection-count metric and the Consumer connection-creation-rate metric to understand how many new connections you are creating. For details, see Broker Metrics and Global Connection Metrics. To reduce usage on this dimension, you can use longer-lived connections to the cluster.
Kafka REST Produce v3	Max throughput: 50 MBps Max active connections: 300 Max connection requests: 300 connection requests per second Max streamed requests: 3000 requests per second	Client applications can connect over the REST API to Produce records directly to the Confluent Cloud cluster. To learn more, see the Kafka REST concepts overview. To learn more about the Kafka REST Produce API streaming mode, see the examples and explanation of streaming mode in the concept docs, and the Produce example in the quick start. A streamed request is a single record to be produced to Kafka. Concatenate the records to produce multiple records in the same request. Delivery reports are concatenated in the same order as the records are sent. For more information, see Produce Records.

Dimensions with a recommended guideline

The following table lists dimensions with a recommended guideline.

These dimensions provide guidelines for capacity planning. The ability to fully utilize these dimensions depend on the workload and utilization of other dimensions. See more about measuring load in cluster load metric and for the maximum bandwidth for each cloud provider (AWS, GCP, Azure), are available in Benchmark Your Dedicated Apache Kafka Cluster on Confluent Cloud.

Dimension	Guideline per CKU	Details
Ingress	~50 megabytes per second (MBps)	Number of bytes that can be produced to the cluster in one second. Available in the Metrics API as received_bytes (convert from bytes to MB). If you are self-managing Kafka, you can look at the producer outgoing-byte-rate metrics and broker kafka.server:type=BrokerTopicMetrics,name=BytesInPerSec metrics to understand your throughput. To reduce usage on this dimension, you can compress your messages. lz4 is recommended for compression. gzip is not recommended because it incurs high overhead on the cluster.
Egress	~150 megabytes per second (MBps)	Number of bytes that can be consumed from the cluster in one second. Available in the the Metrics API as sent_bytes (convert from bytes to MB). If you are self-managing Kafka, you can look at the consumer incoming-byte-rate metrics and broker kafka.server:type=BrokerTopicMetrics,name=BytesOutPerSec to understand your throughput. To reduce usage on this dimension, you can compress your messages and ensure each consumer is only consuming from the topics it requires. lz4 is recommended for compression. gzip is not recommended because it incurs high overhead on the cluster.
Total client connections	~9,000 connections	Number of TCP connections to the cluster that can be open at one time. Available in the Metrics API as active_connection_count. Filter by principal to understand how many connections each application is creating. If you are self-managing Kafka, you can look at the broker kafka.server:type=socket-server-metrics,listener={listener_name},networkProcessor={#},name=connection-count metrics to understand how many connections you are using. This value may not have a 1:1 ratio to connections in Confluent Cloud, depending on the number of brokers, partitions, and applications in your self-managed cluster. How many connections a cluster supports can vary widely based on several factors, including number of producer clients, number of consumer clients, partition keying strategy, produce patterns per client, and consume patterns per client. Confluent derives this guideline from benchmarking that indicates exceeding this number of connections increases produce latency for test clients. However, this doesn’t apply to all workloads. This is why total client connections are a guideline, not a hard limit for Dedicated Kafka clusters. Monitor the impact on cluster load as connection count increases, as this is the final representation of the impact of a given workload or CKU dimension on the cluster’s underlying resources. Consider this guideline a per-CKU guideline. The number of connections tends to increase when you add brokers. In other words, if you significantly exceed the per-CKU guideline, cluster expansion won’t always give your cluster more connection count headroom.
Requests	~15,000 requests per second	Number of client requests to the cluster in one second. Available in the the Metrics API as request_count. If you are self-managing Kafka, you can look at the broker kafka.network:type=RequestMetrics,name=RequestsPerSec,request={Produce FetchConsumer FetchFollower} metrics and client request-rate metrics to understand your request volume. To reduce usage on this dimension, you can adjust producer batching configurations, consumer client batching configurations, and shut down otherwise inactive clients. A high number of requests per second results in increased load on the cluster.

Dedicated limits per cluster

You can add CKUs to a Dedicated cluster to meet the capacity for your high traffic workloads. However, the limits shown in this table will not change as you increase the number of CKUs. For more information about quotas and limits, see Service Quotas for Confluent Cloud and Kafka topic configurations for all Confluent Cloud cluster types.

Dimension	Capability	Additional details
Message size	Max 20 MB	For message size defaults at the topic level, see Kafka topic configurations for all Confluent Cloud cluster types.
Client version	Minimum 0.11.0	None
Request size	Max 100 MB	None
Fetch bytes	Max 55 MB	None
Partitions	Depends on number of CKUs, absolute max 100,000	None
API keys	Default limit 2000	You can request an increase. For more information, see Service Quotas for Confluent Cloud.
Partition creation and deletion	Max 5000 per 5 minute period	The following occurs when partition creation and deletion rate limit is reached: For clients < Kafka 2.7: The cluster always accepts and processes all the partition creates and deletes within a request, and then throttles the connection of the client until the rate of changes is below the quota. For clients >= Kafka 2.7: The cluster only accepts and processes partition creates and deletes up to the quota. All other partition creates and deletes in the request are rejected with a THROTTLING_QUOTA_EXCEEDED error. By default, the admin client will automatically retry on that error until default.api.timeout.ms is reached. When the automatic retry is disabled by the client, the THROTTLING_QUOTA_EXCEEDED error is immediately returned to the client.
Connector tasks per Kafka cluster	Max 250	None
ACLs	Default limit 10000	You can request an increase. For more information, see Service Quotas for Confluent Cloud.

Dedicated limits per partition

You can add CKUs to a Dedicated cluster to meet the capacity for your high traffic workloads. However, the limits shown in this table will not change as you increase the number of CKUs.

The partition capabilities that follow are based on benchmarking and intended as practical guidelines for planning purposes. Performance per partition will vary depending on your individual configuration, and these benchmarks do not guarantee performance.

Dimension	Capability
Ingress per Partition	Max 10 MBps (aggregate producer throughput)
Egress per Partition	Max 30 MBps (aggregate consumer throughput)
Storage per Partition	Unlimited
Storage per Partition for Compacted Topics	Unlimited

Dedicated provisioning time

The estimated time to initially provision a Dedicated cluster depends on the cluster’s size and your choice of cloud provider. Regardless of the cloud provider, sometimes provisioning can take 24 hours or more. Contact Confluent Support if provisioning takes longer than 24 hours.

Note that provisioning time is excluded from the Confluent SLA.

• AWS: the expected provisioning time is one hour per CKU. A dedicated 2 CKU cluster on AWS is estimated to complete provisioning in about two hours.
• GCP: the expected provisioning time is one hour per CKU. A dedicated 2 CKU cluster on GCP is estimated to complete provisioning in two hours.
• Azure: the expected provisioning time is two and a half hours per CKU. A dedicated 2 CKU cluster on Azure is estimated to complete provisioning in five hours.

You will receive an email when provisioning is complete.

Dedicated resizing time

When a cluster is not heavily loaded, expansion and shrink times are generally similar to expected provisioning times (single-digit hours). When a cluster is under heavy load, resizing can take longer.

Resizing a cluster should take about 1-2 hours per CKU.

During a resize operation, your applications may see leader elections, but otherwise performance will not suffer. Supported Kafka clients will gracefully handle these changes. You will receive an email when the resize operation is complete.

Cluster Linking capabilities

Dedicated clusters can be a source or destination of a cluster link, dependent on the networking type and the other cluster involved. To learn more, see Supported cluster types in the Cluster Linking documentation.

Related content

• For per-topic settings and limits, see Manage Kafka Cluster and Topic Configuration Settings in Confluent Cloud
• For quotas that apply to organizations, environments, clusters, and accounts, see Service Quotas for Confluent Cloud
• For performance monitoring your clusters, see Metrics API
• For costs by dimension, see Billing dimensions