Managed Kafka in your cloud or data center
Kafka as a service with the brokers, partitions and replication set for your estate: dedicated clusters in OSStend's cloud, inside your own cloud account, or co-managed in your data center, run 24/7.
45 minutes with an engineer: we review your workload and sketch the design.
- ✓Operated by AceMQ: 160+ customers, 10+ years
- ✓24/7 on-call engineers
An example design file, with illustrative values:
# Kafka in your AWS account service: kafka-events engine: kafka brokers: 6 mode: kraft target: aws/byoc connect: enabled schema_registry: enabled monitoring: included on_call: 24/7
Managed Kafka from OSStend is for teams whose event streaming no longer fits a plan. Engineers design the broker count, partition plan and connectors around your estate, migrate you in with both clusters live, and run Kafka 24/7 in OSStend's cloud, your cloud account or your data center.
- PostgreSQL
- MySQL
- MongoDB
- Redis
- Valkey
- OpenSearch
- Kafka
- RabbitMQ
- AWS
- Google Cloud
- Azure
- Kubernetes
- OpenShift
- Prometheus
- Grafana
- Datadog
- Dynatrace
- Splunk
What OSStend tailors for Kafka
A Kafka cluster is designed from the estate it will carry: topics, partitions, throughput, retention and the systems that produce and consume. The design fixes the broker count, the partition plan, the components around Kafka and where it runs, and you sign it off before anything is built.
- Metrics to your Prometheus or Datadog
- AceMQ engineers on call 24/7
| Area | Set for your workload |
|---|---|
| Size and shape | Broker count, instance size, storage and partition plan, chosen from measured throughput, retention and consumer load. Kafka can add partitions to a topic but cannot reduce them, and adding them changes which partition a key hashes to, so the plan is set with growth in mind. |
| Topology | Multi-zone clusters, multi-region and hybrid links between cloud and data center, and mirroring between clusters for migrations and disaster recovery. For durability, Kafka's documented set-up is replication factor 3, min.insync.replicas of 2 and producers on acks=all; the min.insync.replicas default is 1, so it has to be set on purpose. |
| Configs and components | KRaft-mode clusters, with Kafka Connect and a schema registry included, and the broker and topic settings the workload needs. Kafka 4.0 removed ZooKeeper entirely, and its brokers, Connect and tools need Java 17 while clients need Java 11 (Apache Kafka 4.0 announcement). |
| Versions and upgrades | The Kafka version your clients are validated on, with upgrades planned around your change calendar and never forced. Kafka 4.0 made the KIP-848 consumer group protocol generally available, which targets faster and less disruptive rebalances, and upgrading Java clients to 4.0 needs brokers at 2.1 or higher. |
| Where it runs | OSStend's cloud, your own AWS account, your data center, Kubernetes or OpenShift, or an air-gapped site. |
| Monitoring | Metrics and alerts, including consumer lag and under-replicated partitions, forwarded to your Datadog, Dynatrace, Splunk or Prometheus. |
| Fresh data for AI | Change data capture into Kafka so embeddings update when source data changes. See vector database for RAG and keeping embeddings fresh with Kafka and CDC. |
There is no fixed menu here: the connectors, settings and cluster size your workload needs are written into the design you sign off before anything is built, and that design is what OSStend runs.
Where turnkey Kafka services stop
Turnkey Kafka services come with documented limits. Amazon MSK, for example, caps the broker count per cluster, and past a partition maximum per broker the cluster can no longer be reconfigured or resized.
| Area and managed service | Documented limit | What OSStend does instead | Source |
|---|---|---|---|
| Cluster sizeAmazon MSK | A cluster can have 30 brokers on ZooKeeper and 60 on KRaft. Above the partition maximum per broker, the configuration can no longer be changed and the cluster can no longer be resized. | What OSStend does insteadBroker count and partition plan designed for your estate, and reviewed as topics grow. | AWS docs, checked 30 Oct 2026 ↗ |
Other plan-based services set their own menus of sizes, settings and upgrade timing. The Confluent and Instaclustr alternatives pages compare them with dated sources, and the managed service limits index lists each documented limit with its link.
AWS managed Kafka (Amazon MSK) or Kafka run for you in your own cloud
Amazon MSK, Amazon Managed Streaming for Apache Kafka, is AWS's managed Kafka service and is the natural choice when everything runs on AWS and a self-serve console is what you want. OSStend fits when Kafka must run in another cloud, your own account or a data center, tuned and operated by engineers.
The Amazon MSK comparison sets the two side by side with AWS's published prices and quotas, and says where MSK is the better choice. For the other managed options, see Confluent Cloud and Instaclustr.
Free managed Kafka tiers, and when they stop fitting
As of 15 Nov 2026, Aiven, Confluent Cloud and Redpanda each publish a free or free-to-start managed Kafka tier. Aiven's free service is capped at 250 KB/s and 5 topics. Production teams leave a free tier when sustained throughput, retention, private networking or running in their own cloud account starts to matter.
| Provider | Free offer (as published) | Main limit | Source, checked 15 Nov 2026 |
|---|---|---|---|
| Aiven for Apache Kafka | A free plan at $0 a month, one free Kafka service per organisation, with schema registry and REST proxy included. | Data in and out up to 250 KB/s each, 3 days of retention, up to 5 topics with 2 partitions each. | Aiven free Kafka and Aiven for Kafka |
| Confluent Cloud | The Basic cluster type starts at $0 a month, with the first eCKU free, then $0.14 per eCKU-hour. | Basic is pitched at test cases. The 99.99% uptime SLA is listed from Standard, and private networking from Enterprise. | Confluent Cloud pricing |
| Redpanda Serverless | "Start for free" with no credit card required; Redpanda also offers a 14-day trial of BYOC clusters in your own VPC. | Serverless is multi-tenant on AWS and GCP, with 100 MB/s maximum write throughput and 5,000 partitions on the plan. | Redpanda pricing |
How we compared: each row uses only what the vendor publishes on its own pages, checked 15 Nov 2026, and free tiers change often. Amazon MSK is left out because its pricing page shows no free offer on the day we read it. OSStend is our service.
OSStend has no free tier for Kafka: it is enterprise only. It starts with a 45-minute review with an engineer when a free or plan-based cluster no longer holds the workload. See pricing.
Debezium and Kafka Connect on managed Kafka
Debezium is the usual way to stream database changes into Kafka. OSStend runs Debezium on the Kafka Connect cluster that comes with managed Kafka, or on MSK Connect, and upgrades Debezium and Kafka together.
Managed Debezium covers what is operated 24/7, Debezium vs AWS DMS compares the alternatives, and change data capture explained is the background guide.
Design. Move. Run. for Kafka
For Kafka, an engineer designs the brokers, partitions and replication and you sign it off, OSStend mirrors the topics and owns the staged cutover, and named engineers then run the cluster 24/7.
Design
An engineer reviews your estate (topics, partitions, throughput, retention, consumer groups, connectors and where it must run) and writes the cluster design: broker count, partition plan, KRaft layout, Connect and schema registry. You sign it off before anything is built.
Move
OSStend runs the migration and owns the cutover. Mirroring copies topics from Confluent, MSK, Redpanda or a self-run cluster to the new one while both sides stay live, and producers and consumers move in stages. MirrorMaker 2 can write translated consumer group offsets to the target cluster, but it skips a group that has active consumers there, so each group is moved once its consumers stop on the old side. Traffic switches only when the new side is verified.
Run
Named engineers run it 24/7: broker patching, upgrades on your calendar, partition and capacity changes, consumer lag and replication alerts and incident response, with an SLA written into your contract.
Deployment and service levels
Kafka runs dedicated in OSStend's cloud, inside your own cloud account, or co-managed in your data center. The first two are fully managed; the third is co-managed, with OSStend engineers operating Kafka on infrastructure you own.
Dedicated in OSStend's cloud
Single-tenant Kafka clusters on Google Cloud or AWS, with brokers spread across zones and cross-zone replication traffic counted before sizing.
Fully managedIn your cloud account
Kafka inside your cloud account and VPC, so producers and consumers keep private network paths and the broker bill stays with you. AWS first, then Azure and Google Cloud.
Fully managedIn your data center
Brokers on your hosts, Kubernetes or OpenShift, air-gapped sites included, run through an operator with supported charts and images.
Co-managed
Fully managed Kafka clusters carry an uptime commitment, and a data center deployment carries a 24/7 response-time commitment because the brokers' hosts and disks are yours. The service levels page sets out the tiers, and deployment shows each option in detail.
Uptime and response figures are set per plan and written into your contract. AceMQ, which operates OSStend, sets its support response targets from 15 minutes, by plan.
Proof from AceMQ's operating record
The broker, partition and upgrade work for Kafka is done by AceMQ's engineers. AceMQ operates OSStend with 160+ customers, 10+ years of running open-source infrastructure and engineers on call 24/7, as acemq.com states.
AceMQ names customers only with their written permission, so the Kafka record here stays anonymous. A reference with a similar partition count or migration source can be raised in the 45-minute review.
Pricing for managed Kafka
Managed Kafka is priced from four drivers: the broker design, the storage and retention, the SLA tier and where it runs. An engineer prices the design after a workload review, and there is no free tier or card sign-up for Kafka.
- Broker design
- the replication factor and the Connect and schema registry components set how many brokers and workers have to run.
- Cluster size
- broker count, instance size, storage and retention, which follow partitions and throughput.
- SLA tier
- the uptime commitment on a fully managed cluster, or the 24/7 response-time commitment on a co-managed one.
- Where it runs
- OSStend's cloud, your cloud account or your data center, which decides who pays for brokers and disks.
Give the topic and partition counts, peak throughput, retention and where Kafka must run, and the engineer prices that cluster.
The pricing page shows how the model applies across every engine.
When to contact OSStend, and when not to
Contact OSStend when your Kafka estate no longer fits a plan's broker and partition limits, or has to run in your own account or data center. If a plan still carries it, stay on the plan.
Contact us if
- ✓Your estate has outgrown a plan: more brokers, more partitions or settings the plan does not expose.
- ✓Kafka has to run in your own cloud account or in your data center.
- ✓You are leaving Confluent, MSK or Redpanda and want the migration run for you.
- ✓You want Kafka, PostgreSQL and RabbitMQ under one contract and one on-call team.
Not a fit if
- ✕A standard Kafka plan holds your estate, and it costs less than a designed cluster.
- ✕You want a free tier, or to sign up with a card. For Kafka, OSStend begins with a conversation with an engineer.
- ✕You will keep running Kafka yourself and only want advice: speak to AceMQ.
Questions about managed Kafka
These are the questions architects and platform teams ask most before moving Kafka to a managed service.
What is managed Kafka, and how is it different from Kafka hosting?
Kafka hosting supplies the machines that run brokers. Managed Kafka, also sold as Kafka as a service, adds the operations: broker patching, upgrades, partition and capacity changes, monitoring and 24/7 on-call engineers. OSStend designs the cluster for the estate first, so the service is not a fixed plan.
Can managed Kafka run in my own cloud account?
Yes. The brokers run inside your cloud account and VPC, so your applications keep private network paths and the bill stays with you. AWS comes first, then Azure and Google Cloud. In a data center, Kafka is co-managed on Kubernetes, OpenShift or your own hosts.
How does a move from Confluent, MSK or Redpanda avoid downtime?
Mirroring copies topics to the new cluster while both sides stay live. Producers and consumers move in stages, and traffic switches only after the new side is verified. OSStend engineers own the cutover. The Confluent Cloud comparison page describes the path, and the Confluent to Kafka migration guide covers the phases and the cutover.
What changed in Kafka 4.0 for operators?
Kafka 4.0, announced on 18 March 2025, runs without ZooKeeper in KRaft mode only. Brokers, Connect and tools need Java 17, and clients need Java 11. The KIP-848 consumer group protocol is generally available in it. A cluster still on ZooKeeper needs a planned path to KRaft before it can reach 4.0, and the design review sets that path.
How many replicas does Kafka need to avoid losing data?
Kafka's documentation describes replication factor 3, min.insync.replicas of 2 and producers on acks=all as the typical durable set-up. The min.insync.replicas default is 1, so a topic created with defaults can acknowledge writes held by one replica. The design sets the values per topic class.
Can I reduce the number of partitions later?
No. Kafka documents that it does not support reducing a topic's partition count, and adding partitions changes the key-to-partition mapping for keyed topics, which can affect ordering for existing keys. That is why the partition plan is made with growth in mind, and why a very wrong plan is fixed with a new topic and a migration.
Are Kafka Connect and a schema registry included?
Yes. Both are part of the cluster design, and the connectors and converters the estate uses are listed in the design that you sign off before anything is built.
What does managed Kafka cost?
The price follows the broker count, storage and retention, the SLA tier and where it runs. The pricing page sets out the model, and an engineer prices the specific design after a workload review.
Related pages
- Deployment · OpenShiftKafka on OpenShift and PostgreSQL, run 24/7Kafka on OpenShift and PostgreSQL on OpenShift, co-managed by OSStend through an operator on your cluster, with supported charts and images, 24/7.
- AI data · RetrievalVector database for RAG on your own infrastructureA vector database for RAG on your own infrastructure: when PostgreSQL with pgvector is enough, when OpenSearch hybrid search fits, and how to size retrieval.
- AI data · Fresh contextEmbeddings freshness with Kafka and Postgres CDCPostgres CDC with Debezium and Kafka Connect keeps embeddings current: re-embed on change, handle deletes, and stop agents answering from stale data.
- AlternativeConfluent Cloud pricing explained (2026)Confluent Cloud pricing from Confluent's published rates, read 17 Nov 2026: eCKUs, storage and networking, and when dedicated Kafka is worth pricing.
- Alternative7 Instaclustr alternatives by engine (2026)Seven Instaclustr alternatives for Kafka, PostgreSQL, OpenSearch and ClickHouse, with list prices checked 8 Nov 2026 and where OSStend fits.
- Research · Original dataManaged service limits index, checked 9 Nov 2026A managed service limits index: 13 documented limits of Amazon MQ, MSK, ElastiCache, Neon, Supabase and Instaclustr, each with its source link and date.
Talk to an engineer about your Kafka estate
45 minutes with an engineer: bring your topic and partition counts, your peak throughput and where Kafka runs now, and the engineer sketches the cluster design.
- ✓A design sketch for your workload, not a plan menu
- ✓A zero-downtime cutover path for what you run today
- ✓A price for that design, after the review