HAQM Managed Streaming for Apache Kafka features

HAQM MSK Provisioned provides fine-grained control over your Apache Kafka cluster. You can choose your broker type, pre-provision server instances, select the type of storage you want, and pick the Apache Kafka version of choice. You can also choose when and by how much to scale your clusters in response to workload variation.

HAQM MSK Serverless fully manages your Apache Kafka cluster so you don’t have to estimate how much capacity you need for your workload or decide when to scale it in response to changes in traffic.

No matter which cluster type you choose, with a few steps in the AWS management console, you can create a fully managed cluster that is highly available, secure, and backed by HAQM MSK advanced monitoring and detection systems that automatically maintain the operational health of your cluster.

Apache Kafka uses either Apache Kafka Raft (KRaft) or Apache ZooKeeper for metadata management. HAQM MSK allows you to create clusters in either mode on supported Apache Kafka versions. HAQM MSK also manages these additional metadata nodes for you at no additional cost.

Express brokers are one type of broker offered under MSK Provisioned. Express brokers make Apache Kafka simpler to manage, more cost-effective to run at scale, and more elastic with the same low latency that you expect. Express brokers include virtually unlimited and elastic storage capacity that requires no management overhead, provide up to 3x more throughput per broker, and can scale up to 20x faster. You can also recover up to 90% faster than Standard Apache Kafka brokers in HAQM Managed Streaming services for Apache Kafka (MSK).

Standard brokers under MSK Provisioned offer the most flexibility to configure your cluster's performance. You can choose from a wide range of configurations on the cluster to tune dimensions, including availability, durability, throughput, and latency. On Standard brokers, you also control the storage configurations on your cluster and are responsible for managing storage provisioning and utilization.

Express brokers include virtually unlimited and elastic storage capacity that requires no sizing, provisioning, or on-going capacity management. Storage capacity automatically scales to accommodate your data retention needs and you pay only for the storage you use.

With tiered storage, you can store virtually unlimited data in HAQM MSK without the need to provision and manage storage capacity. You can enable tiered storage in a few steps for existing clusters and only pay for what you use. You can first store data in a performance-optimized primary storage tier, and then let HAQM MSK automatically tier data into the low-cost tier for longer retention. The feature is supported in all AWS Regions where HAQM MSK is available. To learn how to get started with tiered storage, visit the HAQM MSK Developer Guide.

All clusters are distributed across multiple Availability Zones (three is the default), and HAQM MSK offers replicating data across these Availability Zones at no additional cost. Your cluster availability is also backed by the HAQM MSK Service Level Agreement that guarantees three 9s of availability.

HAQM MSK has automated systems that quickly detect and respond to issues. If a component fails, HAQM MSK automatically replaces it without downtime to your applications. We also automatically deploy software patches as needed to keep your cluster up to date and running smoothly.

HAQM MSK Serverless and Express brokers in MSK Provisioned enforce best practice configurations, such as three-way replication, and reserve bandwidth for background operations, such as replication and recovery, so you can more easily achieve predictable availability of your cluster resources.

Using HAQM MSK Replicator, you can set up continual  data replication to a secondary backup cluster in another Region, allowing you to build highly available and fault-tolerant multi-Region applications for increased resiliency. You can also use MSK Replicator to provide lower latency data access in different geographic Regions or to distribute data to your partners.

Your Apache Kafka cluster runs in an HAQM Virtual Private Cloud (HAQM VPC) managed by HAQM MSK. Kafka clients in your own HAQM VPC can privately access the cluster through a cross-account elastic network interface that HAQM MSK deploys in your VPC. If your Kafka clients are spread across one or more VPCs or AWS accounts, you can still privately connect to your cluster by using the multi-VPC private connectivity feature. This feature removes the operational overhead of self-managing an AWS PrivateLink solution and seamlessly scales as the HAQM MSK cluster scales, enabling you to maintain private connectivity to the cluster without making additional configuration changes. Multi-VPC private connectivity also removes the challenges with managing non-overlapping IPs, complex peering and routing tables associated with other VPC connectivity solutions.  The Multi-VPC private connectivity feature allows for overlapping IPs across connecting VPCs.

IAM access control is a no-cost security option that simplifies cluster authentication and Apache Kafka API authorization using IAM roles or user policies to control access. Using IAM access control, you no longer need to build and run one-off access management systems to control client authentication and authorization for Apache Kafka. Your clusters are secured using least-privilege permissions by default. For provisioned clusters, you also can use Simple Authentication and Security Layer (SASL)/Salted Challenge Response Authentication Mechanism (SCRAM) or mutual Transport Layer Security (TLS) authentication with Apache Kafka access control lists (ACLs) to control client access.

HAQM MSK encrypts your data at rest without special configuration or third-party tools. For provisioned clusters, all data at rest can be encrypted using an AWS KMS key by default or your own key. You can also encrypt data in transit through TLS between brokers and between clients and brokers on your cluster. For serverless clusters, all data at rest is encrypted by default using service-managed keys and all data in transit is encrypted by default through TLS.

HAQM MSK offers an option to securely connect to the brokers of HAQM MSK clusters running Apache Kafka 2.6.0 or later versions over the internet. By enabling Public Access, authorized clients external to a private HAQM VPC can stream encrypted data in and out of specific HAQM MSK clusters.

You can scale your MSK Provisioned clusters by adding more brokers or moving to a larger-sized broker instance in minutes with no downtime. Similarly, you can scale down your cluster capacity by removing brokers or by moving to a smaller-sized broker instance.

HAQM MSK Serverless clusters automatically adjust compute and storage resources available to your workloads in response to your application’s throughput needs.

HAQM MSK integrates with Cruise Control, a popular open source tool for Apache Kafka that automatically manages partition assignments on your behalf. For serverless clusters, HAQM MSK automatically manages partition assignments for you.

You can seamlessly scale up the amount of storage provisioned per broker to match storage requirement changes using the AWS Management Console or AWS Command Line Interface (AWS CLI). You can also create an auto scaling policy to automatically expand your storage to meet growing streaming requirements.

With HAQM MSK, you can choose how configurable you want your clusters to be. Express brokers come preconfigured with the HAQM MSK recommended best practice defaults. This provides optimal availability, durability, and throughput performance out of the box. You can customize select configurations to meet the specific needs of your workload. On the other hand, Standard brokers give you the flexibility to modify more than 30 different cluster configurations. This allows you to tailor the availability, price performance, and overall cluster behavior to your exact requirements. You also have access to Kafka's full suite of dynamic and topic-level configurations, helping you to further refine your experience. For more information, see the Custom MSK configurations documentation.

You can visualize and monitor key metrics using CloudWatch to understand and maintain streaming application performance. A default set of over 50 metrics is available at no additional cost. You can also enable more enhanced broker-level and topic-level monitoring to troubleshoot specific issues. Enhanced metrics are billed at standard CloudWatch rates.

Open monitoring with Prometheus lets you monitor HAQM MSK using solutions, such as Datadog, Lenses, New Relic, Sumo Logic, or a Prometheus server, and easily migrate your existing monitoring dashboards to HAQM MSK. For more information, see the Open monitoring with Prometheus documentation.

Broker logs enable you to troubleshoot your Apache Kafka applications and analyze their communications with your MSK cluster. You can deliver Apache Kafka broker logs to one or more of the following destination types: HAQM CloudWatch Logs, HAQM Simple Storage Service (HAQM S3), and HAQM Data Firehose. You can also log HAQM MSK API calls with AWS CloudTrail.

We offer a wide variety of AWS integrations in HAQM MSK. These integrations include the following:

•  AWS Identity and Access Management (IAM) for Apache Kafka and service-level API access control
• HAQM Managed Service for Apache Flink for running fully managed Apache Flink applications to process streaming data within Apache Kafka
• HAQM Managed Service for Apache Flink Studio for running interactive streaming SQL and long-running SQL jobs using Apache Flink SQL
• AWS Glue Schema Registry for centrally controlling and evolving schemas
• AWS IoT Core for IoT event streaming into HAQM MSK
• AWS Database Migration Service (AWS DMS) for change data capture and analytics
• HAQM VPC for private client connectivity and network isolation
• AWS KMS for at-rest encryption
• AWS Certificate Manager Private Certificate Authority for mutual TLS client authentication
• AWS Secrets Manager for secure storage and management of SASL/SCRAM secrets
• AWS CloudFormation for deploying HAQM MSK in code
• HAQM CloudWatch for cluster-, broker-, topic-, consumer-, and partition-level metrics