Migrate to ClickHouse

Migrate from BigQuery to ClickHouse

BigQuery charges by the byte scanned and is tied to one cloud. ClickHouse gives you predictable cost, sub second latency at high concurrency, and the freedom to run on GCP, anywhere else or on premise.

How the migration works
Sub second
Queries over billions of rows, from columnar storage and vectorised execution.
50%+
Cost savings observed in the migrations we have delivered from SQL Server, OpenSearch and Snowflake.
100s
Of concurrent users served with consistent latency, so analytics can go into your product.
Anywhere
Open source. Run it on any cloud, on premise or as ClickHouse Cloud.
Why teams move

Why BigQuery users come to us

BigQuery is a capable warehouse for large batch queries, and if you are entirely on Google Cloud it is the path of least resistance. The problems appear when it starts backing anything interactive. On demand pricing bills for every byte scanned, so a popular dashboard or an exploratory analyst becomes a cost event, and slot reservations trade that for a fixed bill that is often larger.

Latency is the other issue. BigQuery queries carry a fixed overhead of a second or more before any data is read, which is fine for a scheduled report and not fine for a product with users waiting on a screen. It is not designed to serve hundreds of concurrent low latency queries, and you would not use it as an application backend in the way you would ClickHouse.

ClickHouse is open source and runs anywhere, including on GCP next to your existing data, so a migration does not have to mean leaving Google. It means moving the real time, interactive and customer facing analytics onto an engine built for them, with a cost model you can predict.

Sound familiar?
  • [01]Monthly bills vary with query volume, and a single badly written query can cost real money.
  • [02]Interactive dashboards feel slow because every query carries a second or more of overhead.
  • [03]Streaming inserts, slot reservations and BI Engine are being paid for to work around latency.
  • [04]Serving analytics into a product is impractical on cost and concurrency.
  • [05]Everything is tied to Google Cloud, with no option to run elsewhere or on premise.
What changes

BigQuery to ClickHouse: faster, cheaper, simpler

ClickHouse is the fastest open source analytical database on the market. These are the differences that matter most for teams coming from this platform.

[01]

Predictable, lower cost

ClickHouse bills for compute and storage, not bytes scanned. Heavy compression and efficient execution mean the compute you need is modest, and ClickHouse Cloud scales it linearly in small steps.

[02]

Milliseconds, not seconds

There is no query startup overhead. Point lookups and aggregations over billions of rows return in milliseconds, which is what interactive dashboards and embedded analytics need.

[03]

Hundreds of concurrent users

ClickHouse is designed to be queried by applications and many users at once, so analytics can be offered to customers rather than kept internal.

[04]

Not tied to a cloud

Run ClickHouse on GCP, AWS, Azure, on your own hardware or as ClickHouse Cloud. BigQuery only runs in one place; ClickHouse runs wherever you need it.

How we migrate

Accelerated migration with CDC

BigQuery migrations are export driven. We export tables to Parquet in GCS, which ClickHouse reads directly with its table functions or ingests continuously with ClickPipes. The upstream sources that currently feed BigQuery, such as Pub/Sub, Kafka or operational databases, are then connected directly to ClickHouse. BigQuery SQL, scheduled queries and Looker models are translated as part of the engagement.

  1. 01

    Assess

    We review the schemas, the query workload and the cost of the current platform, then agree the scope, the target architecture and the business case before anything is built.

  2. 02

    Design the ClickHouse schema

    Tables are redesigned for columnar storage rather than copied across one to one. Sorting keys, partitioning, codecs and materialised views are chosen against your real queries.

  3. 03

    Load and replicate

    Historical data is bulk loaded, then changes are streamed continuously so ClickHouse stays in line with the source for as long as the two need to run side by side.

  4. 04

    Run in parallel

    Dashboards, applications and reports are pointed at ClickHouse and validated against the source. Performance and cost are measured rather than assumed.

  5. 05

    Cut over

    Workloads move across in stages, the old platform is scaled down and the remaining spend is retired. Nothing is switched off until the numbers reconcile.

  6. 06

    Support

    After go live we can monitor the environment, tune it as the workload grows and provide an escalation path under SLA through our proactive support service.

Tooling we use
Parquet and object storage

Warehouse exports land in S3 or GCS as Parquet or Iceberg. ClickHouse reads them directly with its table functions, or ClickPipes keeps ingesting new files as they arrive.

ClickPipes

The managed ingestion service built into ClickHouse Cloud. Its CDC connectors take an initial snapshot and then stream every insert, update and delete, with no infrastructure for you to run.

We are tool agnostic. The right choice depends on whether you are going to ClickHouse Cloud or self hosted, how much control you want over the pipeline and what else is in the stack. Zero downtime is the norm, not the exception.

Beyond the migration

One database, many use cases

Moving off BigQuery is usually the start. Because ClickHouse is so versatile, the same platform tends to pick up more workloads once it is in place.

Real time analytics

Dashboards and embedded analytics over data that is seconds old, served to many users at once.

Data warehousing

A warehouse that is fast enough to query directly, without the overnight batch or the cube layer.

Observability

Logs, metrics and traces at a fraction of the cost of Datadog, Splunk or Elastic, with ClickStack on top.

Time series

Market data, IoT, telemetry and events at hundreds of thousands of rows per second with long retention.

Search and security analytics

Full text search, SIEM and threat hunting over high volume event data using plain SQL.

Agentic AI

A data platform fast and cheap enough for agents that turn one question into dozens of queries.

Related work

Migrations we have delivered

Gaming

Helping An Online Gaming Company Migrate From Postgres To ClickHouse Cloud

Building a real-time data warehouse for a sports betting and gaming platform processing ~300 million events per day, scaling to trillions of rows.

Crypto

Using ClickHouse To Power Crypto Analytics For The Leading Hyperliquid Blockchain

Delivering real-time analytics based on blockchain data from the Hyperliquid network using ClickHouse materialized views and an innovative API layer.

Further reading

Ready to move off BigQuery?

Start with a short call. We will look at your schemas, queries and current spend, tell you honestly whether ClickHouse is the right fit, and put a number on the performance and cost you can expect.

Support after go live