Aiven

Editor’s note: This Aiven review is based on my hands-on evaluation of Aiven for ClickHouse 26.3 on the Aiven console in October 2026. I weighed pricing, setup, Kafka and PostgreSQL integrations, storage and backups, SQL limits and versions, and rivals. Aiven runs Kafka, PostgreSQL, OpenSearch, Valkey and more, and I reviewed it through its ClickHouse service.

Quick verdict: I recommend Aiven for teams that already run (or plan to run) Kafka or PostgreSQL on Aiven and want ClickHouse on a fixed monthly bill in their preferred cloud. The trade-off: one daily snapshot, no point-in-time recovery, and compute only grows by switching plans.

Key Takeaways:

• Startup-8 (1 node, 2 vCPU, 8 GB) costs about $180/month and three-node Business-8 about $500/month on AWS us-east-1.
• Kafka and PostgreSQL integrations come from the console, up to 400 Kafka-engine tables per integration.
• ClickHouse has no free plan, only the 30-day, no-card trial.
• Backups run once a day with no point-in-time recovery, and Startup is a single node with no replica.
• It runs on AWS, Google Cloud and Azure only; 26.3 is the default version.

In this review, I’ll take a closer look at Aiven’s pricing, integrations, storage and SQL limits, so you can see exactly whether it beats ClickHouse Cloud, Altinity.Cloud or Tinybird for your workload.

Pros and Cons

How Much Does Aiven for ClickHouse Cost?

Aiven bills ClickHouse by the hour per plan, all nodes included, with VMs, network and backups in the price. The cheapest plan, Startup-8, is $0.24658/hour on AWS us-east-1, about $180/month. Every price below is AWS us-east-1, because rates change by cloud and region.

• Startup (from about $180/month): for test and development environments with high performance needs, on one node.
• Business (from about $500/month): for production, with three nodes across three availability zones.
• Premium (from about $3,800/month): for sharded workloads, with 6 to 30 nodes in shards of three.
• BYOC (quote only): for running in your own AWS, Google Cloud or Azure account, with a commitment deal and the Advanced or Premium support tier.

AWS us-east-1, USD. Monthly = hourly rate x 730 hours; hourly figures are rounded.

Plan	Nodes	Per node	Hourly	About per month	Backups kept
Startup-8	1	2 vCPU / 8 GB	$0.247	$180	3 daily
Startup-16	1	4 vCPU / 16 GB	$0.411	$300	3 daily
Startup-64	1	16 vCPU / 64 GB	$1.301	$950	3 daily
Business-8	3	2 vCPU / 8 GB	$0.685	$500	15 daily
Business-16	3	4 vCPU / 16 GB	$1.370	$1,000	15 daily
Business-64	3	16 vCPU / 64 GB	$5.479	$4,000	15 daily
Premium-6x-16	6	2 vCPU / 16 GB	$5.205	$3,800	30 daily
Premium-30x-64	30	8 vCPU / 64 GB	$104.18	$76,050	30 daily

• Extra disk: about $0.20/GB-month in 50 GB steps; Startup and Business include 100 GB.
• Tiered object storage: about $0.14/GB-month.
• PrivateLink: billed extra, and your cloud provider may charge for traffic to other cloud services.
• Billing: hourly while powered on, one-hour minimum; powering off stops the charges.
• Support: Basic is included. Essential (24/5), Advanced (24/7, 6-month minimum) and Premium (24/7, 12-month minimum) have no public price.

The 30-day trial needs no card and allows up to 10 VMs. ClickHouse is not on Aiven’s free plan list, which covers Kafka, MySQL, OpenSearch, PostgreSQL and Valkey.

The cheapest cloud flips by plan: Startup-8 is $0.199/hour on Google Cloud us-east1, but Business-8 is cheapest on AWS ($0.685 vs $0.747 on Google and $0.832 on Azure eastus). Price the production plan in the region you actually need.

Is Aiven for ClickHouse Good Value for Money?

• Predictable bill: a fixed plan beats ClickHouse Cloud’s usage billing ($0.2181 per compute unit-hour on Basic plus $25.30/TB-month storage) for steady workloads I can forecast.
• Steep production jump: Business-8 costs about 2.8x Startup-8 for three replicas of the same data, not three times the storage.
• Cheaper BYOC route: Altinity.Cloud BYOC starts at $0.0625 per vCPU-hour without a commitment deal.
• Small API projects: Tinybird’s $0 plan with 10 GB is the cheaper start for analytics APIs.

I recommend Business-8 for customer-facing analytics and Startup-8 for schema work. Premium only pays off when one 3-node shard can no longer hold the data.

Reviewer’s Notes: Use the trial on Startup-8 to size the schema, then go straight to Business-8 for anything a customer sees. Startup’s lower price buys one node with no replica, and the 99.99% uptime objective on its SLA does not replace a second copy of the data.

My Experience With Aiven

The console path from signup to a Kafka-fed ClickHouse table is fully point-and-click, with no infrastructure scripts.

Creating the Service

1. I opened Services > Create service > ClickHouse.
2. I picked the cloud (AWS, Google Cloud or Azure), then the region and the plan.
3. I named the service, added optional tags and clicked Create service.
4. The service showed Rebuilding, then switched to Running once ready.

The trial credits covered the ClickHouse service without a card on file. The same service can be created with:

• The avn CLI.
• The Terraform provider.
• The Aiven Operator for Kubernetes.
• The Aiven API.

Connecting and Running the First Query

My first SQL ran in Data > Query editor, which executes as the default user. The ClickHouse /play UI also works over HTTPS. For applications, Aiven supports:

• Drivers for Python, Go, Java, Node.js and PHP, plus JDBC.
• clickhouse-client over native TCP.
• The MySQL interface, and Arrow Flight on new 25.8 and 26.3 services.

TLS is mandatory: plain HTTP, gRPC and the PostgreSQL interface are not offered, so a self-hosted app that connects without TLS needs its connection settings changed. Users and roles live in the console’s Users tab or in SQL (CREATE USER, GRANT), and I give each application its own user rather than reuse the query editor’s default.

Wiring In a Kafka Topic

The Kafka setup starts at Data > Integrations > Apache Kafka, pointing at an Aiven for Kafka service or an external Kafka endpoint. The integration asks for:

• A table name and consumer group.
• The topics and data format.
• Column names and types.

The new table is queried as <kafka_service>.<table>. For storage, I land it in a MergeTree table through a materialized view, because the Kafka-engine table only reads the stream. Schema Registry works only with Aiven for Kafka, not external Kafka, which matters for pipelines that rely on registered schemas.

Changing the Plan

Plan changes run from the console, Terraform or avn service update --plan. Nodes are recycled and rebuilt while the service stays reachable, and the new plan needs at least 125% of the current disk.

Reviewer’s Notes: The friction is not setup, it is the SQL restrictions. Aiven forces Replicated engines and lets only avnadmin run CREATE DATABASE, so check an existing schema against those rules before the first production import.

Kafka, PostgreSQL and Data Lake Integrations

On Startup plans and up, Aiven for ClickHouse takes data source integrations from Aiven for Kafka (or external Kafka), PostgreSQL, MySQL, another ClickHouse service, Amazon S3 and Azure Blob Storage, with up to 400 Kafka-engine tables per integration.

This is the strongest reason to pick Aiven: Kafka, PostgreSQL and ClickHouse run in one console and one bill. ClickHouse Cloud and Altinity.Cloud host ClickHouse, but neither runs your Kafka and PostgreSQL services next to it.

📨 Kafka Engine Through the Integration

Aiven allows the Kafka engine only through an integration, never through SQL CREATE TABLE. External Kafka endpoints work, but Schema Registry support needs Aiven for Kafka. A separate managed-databases integration also exists for Kafka and PostgreSQL. Aiven pitches this pairing for streaming millions of events per second from Aiven for Kafka into ClickHouse, with Grafana dashboards on top. The integration handles the connection; the destination MergeTree schema and materialized view are still mine to design.

🐘 PostgreSQL and MySQL Federation

The PostgreSQL engine works through SQL or an integration, and managed credentials store connection details as named collections for PostgreSQL, MySQL, ClickHouse, S3 and Azure Blob. The pattern is OLTP in PostgreSQL, OLAP in ClickHouse: transactional writes stay in Aiven for PostgreSQL, and federated queries read those tables from ClickHouse without keeping a local copy. MySQL sources follow the same model through the MySQL engine.

🪣 S3, Iceberg, Delta Lake and Hudi

Federated queries are enabled by default, with these engines:

• S3 and S3Queue (S3Queue, once a request on Aiven’s ideas forum, is now supported).
• Iceberg, DeltaLake and Hudi.
• URL and AzureBlobStorage.

These engines cover querying lake tables in place and ingesting files from S3-compatible storage. Reading a bucket is not the same as storing in it: you cannot attach your own bucket as the tiered-storage layer.

📊 Monitoring With Grafana

Metrics push to Aiven for Metrics (Thanos) or Aiven for PostgreSQL, with dashboards in Aiven for Grafana. Datadog and Prometheus endpoints cover teams whose monitoring already lives outside Aiven. Better slow-query insight is still an open request on Aiven’s ideas forum, so plan on building those dashboards yourself.

Integration	How it connects	Plan needed	Catch
Kafka	Kafka engine created by the integration	Startup or higher	Schema Registry only with Aiven for Kafka
PostgreSQL	PostgreSQL engine via SQL or integration	Startup or higher	Federated reads, not a local copy
MySQL	MySQL engine with managed credentials	Startup or higher	Federated reads, not a local copy
S3 / Azure Blob	Object-storage engines	Startup or higher	Your bucket cannot become the storage tier
Grafana / metrics	Aiven for Metrics or PostgreSQL, then Grafana	Separate Aiven services	Datadog and Prometheus as outside options

Storage, Backups and High Availability

Aiven takes one snapshot a day (plus one before power-off) and keeps 3 on Startup, 15 on Business and 30 on Premium. There is no point-in-time recovery: restoring means forking from a daily backup, so anything inserted since the last snapshot is lost if every node in a shard fails.

The cluster is multi-master. ClickHouse replicates the data, ZooKeeper (one per node) replicates schema and users, and one service URL routes connections to random nodes. Business and Premium replicate each shard across three availability zones, but a region loss is not covered and there is no cross-region backup.

Tiered Storage

Tiered storage pairs block storage with object storage:

• AWS: EBS gp3.
• Azure: Premium SSD v2.
• Google Cloud: Hyperdisk Balanced.

Parts move to object storage once local disk passes 80% full or after a time threshold, a local cache for remote files is on by default, and backups cover both tiers. Object storage costs $0.14 vs $0.20 per GB-month for block disk on AWS us-east-1. Once enabled, tiered storage cannot be turned off (the workaround is copying data to a new table), and it is not available on BYOC.

Scaling Disk and Compute

The disk autoscaler grows disk at 85% usage by at least 10% and never shrinks it. Compute changes only through a plan change. Because every node in a shard holds the same data, total storage is bounded by what one node holds, not the sum of three.

Aiven commits to a 99.99% monthly uptime objective with service credits on Startup, Business and Premium, but maintenance can cause up to one hour of downtime, and Startup has no replica to fail over to. For contrast, ClickHouse Cloud Basic runs in a single availability zone with 1-day backup retention; its Scale tier adds two or more zones.

Reviewer’s Notes: Keep raw events in Kafka or S3 long enough to replay at least a day. Replicas protect availability, but the daily snapshot is the only restore point if a whole shard is lost.

SQL Limits, Versions and Upgrades

Aiven remaps every MergeTree-family table to its Replicated variant, does not support the Log engine (Buffer is the suggested substitute), and lets only avnadmin run CREATE DATABASE, using the Replicated database engine. DDL then replicates without ON CLUSTER. Buffer is not a drop-in Log replacement, so check engine behavior, not just whether a statement parses.

Other limits that shape capacity and schema design:

• Concurrent queries: 100 per 16 GB RAM, within a range of 25 to 400.
• Connections: 4,000 per node.
• Databases: up to 400 per service.
• Premium sharding: Distributed tables are required to query all shards.
• SharedMergeTree: ClickHouse Cloud’s engine is not available.

Version Policy

Aiven tracks upstream LTS releases (.3 in March, .8 in August), supports two at a time and reaches GA about four months after upstream. 26.3 is the default for new services, and 25.8 is still selectable. 25.3 reached end of life on September 30, 2026 and was auto-upgraded to 25.8; new 25.8 services can be created until November 30, 2026, with end of life on February 28, 2027. Critical patches not applied within 14 days are applied automatically.

Upgrading to 26.3

26.3 brings GA text indexes for full-text search and turns async_insert on by default. The upgrade is blocked by:

• Object(‘json’) columns.
• LiveView tables.
• The removed DEFLATE_QPL and ZSTD_QAT codecs.
• 25.8 text indexes, which must be dropped first.

Services still on 25.3 go through 25.8 first, and there is no downgrade, so test on a fork. Set SET async_insert = 0 only if you need synchronous inserts. If your schema depends on the Log engine or SQL-created Kafka tables, self-hosted ClickHouse is the route that keeps them.

Security, Networking and BYOC

TLS is required on every interface, and Aiven for ClickHouse supports IP allow-lists, private access in a VPC, PrivateLink (billed extra) and static IPs, all toggled in the service’s advanced parameters.

• VPC peering: the trial allows 1 VPC and 10 peering connections, enough to test a private-network layout.
• Users and roles: managed in the console or with SQL CREATE USER and GRANT, so application and admin access stay separate.
• BYOC: runs in your own AWS, Google Cloud or Azure account (OCI on request), but needs a commitment deal plus Advanced or Premium support, and pricing goes through sales.
• Tiered storage on BYOC: it works differently from the Aiven-hosted version, so settle the storage design in the BYOC proposal.

Altinity.Cloud is the self-service alternative here: it publishes BYOC rates from $0.0625 per vCPU-hour and offers a 14-day BYOC trial. I shortlist it first when owning the cloud account is mandatory and a negotiated Aiven commitment is not. Aiven’s BYOC fits an organization consolidating several managed data services, not a small team wanting ClickHouse alone in its own account.

How Does Aiven Compare to Competitors?

Aiven wins on fixed plans plus Kafka and PostgreSQL in the same console and bill. For ClickHouse alone, each rival wins a clear niche:

• ClickHouse Cloud: the first-party service, with SharedMergeTree and, on Scale, compute-compute separation and automatic vertical scaling. Compute runs $0.2181 per unit-hour on Basic, $0.2985 on Scale and $0.3903 on Enterprise, plus storage at $25.30 per TB-month (N. Virginia), with a 30-day trial. Basic caps out at 1 TB storage and 8-12 GiB memory in one availability zone. It wins for spiky workloads and the newest features, though compute-unit billing is harder to forecast than a fixed plan; see the directory’s ClickHouse entry.
• Altinity.Cloud: the ClickHouse specialist. BYOC runs from $0.0625 per vCPU-hour (Dev) through $0.0833 (Enterprise) to $0.1111 (Premium), storage carries no BYOC fee, and it runs on AWS, Google Cloud, Azure and Hetzner with a 14-day BYOC trial. That works out to about $45, $60 and $80 per vCPU-month, but hosted (non-BYOC) rates come only from a calculator, and there is no Kafka or PostgreSQL service family around it. It wins when ClickHouse must live in your own account.
• Tinybird: a $0 plan with 10 GB and a Developer plan at about $25-49/month that publishes SQL as API endpoints, but you never get a raw ClickHouse server, and compute is shared on lower plans. It wins for product teams shipping analytics APIs.
• Instaclustr for ClickHouse (NetApp): tiered storage at no extra charge and a 99.99% SLA on enterprise deployments, but no public prices, and its product pages list AWS and Azure only. It wins for teams already on Instaclustr Kafka or Cassandra.

Option	Pricing model	Entry point	Runs in your cloud	Best for
Aiven	Fixed hourly plans	About $180/month (Startup-8)	BYOC with commitment deal	Teams on Aiven Kafka/PostgreSQL
ClickHouse Cloud	Compute units plus storage	$0.2181/unit-hour (Basic)	Hosted by ClickHouse	Spiky workloads, newest features
Altinity.Cloud	Per vCPU-hour (BYOC)	$0.0625/vCPU-hour (Dev)	Yes, 14-day BYOC trial	ClickHouse in your own account
Tinybird	Plans plus usage	$0 with 10 GB	Hosted platform	Analytics APIs
Instaclustr	Quote only	Contact sales	Yes	Existing Instaclustr customers

DoubleCloud, once a managed ClickHouse option, stopped onboarding on October 1, 2024 and shut down on March 1, 2025.

For an existing Aiven customer, keeping the ingestion source and the analytics database together outweighs ClickHouse-only features. For a new ClickHouse-only project, I compare ClickHouse Cloud and Altinity.Cloud first.

How I Reviewed Aiven

I evaluated Aiven for ClickHouse 26.3 as a production analytics service, with extra weight on recovery limits and the cost of moving past a single node. I covered:

• Service creation on the trial and cloud selection.
• The console query editor and the /play UI.
• The Kafka and PostgreSQL integration flows.
• Plan changes, tiered storage and backup retention.
• SQL engine rules and version transitions.

I compared Aiven’s per-plan hourly rates across AWS, Google Cloud and Azure in US East, then against ClickHouse Cloud, Altinity.Cloud, Tinybird and Instaclustr. I also weighed what users keep asking for on Aiven’s ideas forum (more frequent backups, cross-region replication, clearer error messages) and Aiven’s status history, including a June 2024 incident where a wrong intermediate TLS certificate broke ClickHouse connections. I ran no benchmarks, so this review makes no speed claims.

Prices current as of October 2026.

Should You Use Aiven for Managed ClickHouse?

Among real-time analytics databases in the directory, Aiven for ClickHouse stands out for fixed, all-inclusive plans and native Kafka and PostgreSQL integrations. Against that sit daily-only backups, plan-step scaling and a locked-down SQL surface.

Buy Aiven if:

• You already run Kafka or PostgreSQL on Aiven, so the integrations work from day one.
• You want a fixed monthly bill, such as Business-8 at about $500/month on AWS us-east-1 with three replicas.
• You need a choice of AWS, Google Cloud or Azure across about 90 regions.

Skip Aiven if:

• You need point-in-time recovery or cross-region backup.
• Your load is spiky; ClickHouse Cloud’s usage billing fits better.
• You need ClickHouse in your own account without a commitment deal; Altinity.Cloud BYOC does that.
• You only need analytics APIs; Tinybird is cheaper to start.
• Your schema depends on the Log engine or SQL-created Kafka tables; self-hosted ClickHouse keeps them.

If you buy, budget for replayable source data in Kafka or S3, because Business replication covers an availability-zone failure but not the gap between daily snapshots.

Start the 30-day trial on Startup-8, wire one Kafka topic through the integration, and test your DDL against the Replicated-engine rules before committing to Business.

Aiven FAQ

How much does Aiven for ClickHouse cost?

Aiven for ClickHouse starts at about $180/month for Startup-8 (1 node, 2 vCPU, 8 GB) on AWS us-east-1. Business-8, with three nodes, costs about $500/month, and Premium starts around $3,800/month for six nodes. Monthly figures assume 730 powered-on hours; region, extra disk and PrivateLink change the bill.

Does Aiven have a free plan for ClickHouse?

No. Aiven has no free ClickHouse plan. The 30-day trial needs no credit card, allows up to 10 VMs and its credits can be spent on ClickHouse. Aiven’s free plans cover Kafka, MySQL, OpenSearch, PostgreSQL and Valkey only. Services power off at the end of the trial unless a payment card is added.

Is Aiven for ClickHouse the same as ClickHouse Cloud?

No. Aiven runs open source ClickHouse on Aiven-managed VMs with fixed plans and Replicated engines. ClickHouse Cloud is the first-party service, with SharedMergeTree and usage-based compute billing from $0.2181 per unit-hour on Basic. Pick Aiven for the shared Kafka and PostgreSQL platform, and ClickHouse Cloud for elastic capacity.

Can I restore Aiven for ClickHouse to a point in time?

No. Aiven for ClickHouse has no point-in-time recovery. It takes one snapshot a day and keeps 3 on Startup, 15 on Business and 30 on Premium. Restoring creates a fork from one of those backups, so keep replayable events in Kafka or S3 to cover data inserted after the latest snapshot.

Which ClickHouse versions does Aiven support?

Aiven supports 26.3 as the default and 25.8 as a selectable version. 25.3 reached end of life on September 30, 2026, and services were auto-upgraded to 25.8. New 25.8 services can be created until November 30, 2026, with end of life on February 28, 2027. Downgrades are not supported, so test upgrades on a fork.