ClickHouse announces Series C extension and key leadership hires ->->

Comparisons
/
Observability

ClickHouse vs Elastic

ClickStack is a high-performance, open-source observability stack built on ClickHouse. It delivers high compression, lightning-fast queries and powerful aggregations across high cardinality logs, metrics, traces, session replays at petabyte scale.

Elastic, by contrast, is rooted in a full-text search engine, built on the belief that observability was just a search problem at a time when inverted indices were sufficient at smaller scales. Never designed for metrics, it cannot unify logs, metrics, and traces in a single system, leaving observability fragmented, costly, and slow.

Why ClickStack is better:

Reduction in costs

10x

Faster analytical queries

Better compression

Read our comprehensive guide about migrating from Elastic to ClickStack.

Frustrated by slow queries, rising storage costs and an endless need to scale horizontally? You’re not alone.

1 Queries executed

Join others migrating to ClickStack from Elastic

Extensive testing and comparison showed that ClickHouse outperformed other solutions like Elasticsearch in three key areas: performance, compatibility, and cost-efficiency for large-scale operations.

Elasticsearch was fast but prohibitively, laughably expensive. With ClickHouse, queries that once took seconds came back in milliseconds, logs were searchable within 20 seconds, and the system finally felt absurdly fast

ClickHouse logging cluster has now exceeded 400 physical nodes, with a peak writing traffic of over 40 GB/s. This supports approximately 15 million queries per day, with a peak QPS of about 200. Compared to Elasticsearch, the machine cost of ClickHouse has decreased by 30%. The query speed has improved by about 4 times compared to Elasticsearch

In the past, we stored all of these logs on self-hosted Elasticsearch. Clickhouse emerged to be the best candidate for our use case, and we decided to move forward with it.

Performance at Scale

Elastic slows under heavy ingest and high-cardinality queries, while ClickHouse powers sub-second analytics even at petabyte scale.

Full data set aggregation for 1 billion JSON documents

Lower Cost, Higher Efficiency

ClickHouse’s columnar storage and advanced compression cut storage needs by > 50%, reducing infrastructure costs dramatically and allowing for long term retention.

Storage required for 1 billion JSON documents

Unified Observability

ClickStack runs logs, metrics, and traces in one engine alongside business and application data for unrivalled correlation. Elastic was never designed for analytical workloads leaving data fragmented.

Open source and open standards

ClickStack is fully open source (MIT + Apache 2.0) and OpenTelemetry-native, ensuring interoperability and freedom from lock-in.

ClickStack compared to Elastic Observability

At a high level, Elastic and ClickStack share a familiar shape: both have a data collection layer (Beats and Logstash vs. OpenTelemetry), a storage engine (Elasticsearch vs. ClickHouse), and a UI (Kibana vs. HyperDX). But beneath these parallels, the architectures diverge.

Elastic is a distributed search engine built around inverted indices and a shard-based architecture. While effective for full-text search, this design introduces high storage overhead, limited query parallelization, and heavy contention between ingest and query workloads.

ClickStack is powered by ClickHouse, a database built on a columnar, shared-nothing architecture optimized for analytics. It minimizes storage with advanced compression, parallelizes queries across all available cores, and separates storage from compute in the cloud to deliver fast, efficient observability at scale. Full SQL support unlocks deep data analysis.

Elastic’s scalability is constrained by shards and the JVM, built for terabyte workloads. Shards are size-limited, queries parallelize only within shard boundaries, and JVM heap caps (~64GB) force horizontal sprawl. This drives network overhead, costly rebalances on outages, and resource contention between ingest and queries, often pushing users to sample data.

ClickStack, powered by ClickHouse, scales to petabytes. Queries parallelize across all cores without JVM limits, and vertical scaling reduces reliance on shards. In ClickHouse Cloud, compute and storage scale independently, all nodes share a single S3 copy for cost-efficient retention, and reads/writes can be isolated with elastic warehouses.

ClickStack, built on ClickHouse, is engineered for speed. Its vectorized engine, columnar compression, and efficient memory use deliver sub-second aggregations on high-cardinality data at petabyte scale - critical for observability workloads. Elastic, designed for search, struggles with analytics: aggregations are slow at scale, term aggregations approximate, and indexing competes with queries for resources in open-source.

Benchmarks and user reports show ClickHouse outperforming Elasticsearch by 10–100x on analytics with at least 2x better compression. Without tuning or schema changes, it handles complex aggregations and joins across billions of rows at a fraction of the time and cost.

ClickStack is OpenTelemetry-native but not exclusive, supporting OTel out of the box while also accommodating wide events and other flexible models. Elastic, by contrast, appends OTel support to its stack, with historical agents still migrating.

Native table engines read directly from Postgres, MongoDB, and object storage, with Parquet and Lake formats supported natively with external catalogs like Hive and Glue for querying data in place - so teams can build observability on open standards without lock-in or data duplication. ClickPipes extends this with hosted ingestion from S3 and Kafka, removing the need to manage pipelines.

Yes
Native object storage support
Yes
Aggregate states for accuracy
Yes
Columnar storage with high compression
Yes
ClickPipes for streaming ingest
Yes
JSON with type fidelity
Yes
1,000+ QPS per node
Yes
Full parallelization within & across shards
Yes
Full join support
Yes
Materialized views (incremental & refreshable)
Yes
Full SQL support
Yes
Natural language search via HyperDX
Yes
Self-managed & cloud
Yes
Decoupled compute/storage (ClickHouse Cloud)
Yes
Stateless compute nodes (ClickHouse Cloud)
Yes
Real-time ingest supported
Yes
Async inserts for small batches
Yes
Inverted index support
Yes
REST API support
Yes
Incremental materialized views
Yes
Unlimited shard sizes (disk-based)
Yes
Parallel replicas distribute queries
Yes
Vertical scaling to very large machines
Yes
Up to 1,000 concurrent queries per node
Yes
Horizontal scaling with sharding
Yes
Read/write isolation (ClickHouse Cloud)
Yes
Elastic scaling with load (ClickHouse Cloud)
Yes
Custom hardware profiles available
Yes
Granular cache control
Yes
Efficient joins across billions of rows
Yes
High compression built-in
Yes
Accurate aggregations at scale
Yes
1,000+ concurrent queries per node
Yes
Materialized views (incremental & refreshable)
Yes
Tens of millions/second
Yes
70+ formats including Parquet, ORC, Avro, JSON, CSV
Yes
70+ formats including Parquet, ORC, Avro, JSON, CSV
Yes
Query in-place via table engines (e.g. Postgres, S3)
Yes
Hosted ingest from Kafka & S3
Yes
Open format support
Yes
Third-party catalog support
Yes
Connect to Postgres, MongoDB, MySQL, S3, Kafka, and more
Yes
Supported

No
Paid, Elastic Cloud Serverless only
No
Terms aggregations are estimates
Intermediate—
Doc values provide columnar storage, not compressed
Intermediate—
Requires Logstash/third parties, no hosted ingestion
Intermediate—
First event field determines type
Intermediate—
Requires horizontal scaling/replicas
Intermediate—
Limited; relies only on shards
Intermediate—
LOOKUP JOIN only
Intermediate—
Limited types; requires full scans
Intermediate—
Limited syntax coverage
Yes
Natural language search supported
Yes
Self-managed & Elastic Cloud
Yes
Elastic Cloud Serverless
Yes
Elastic Cloud Serverless
Yes
Real-time ingest supported
Yes
Inserts supported
Yes
Inverted index supported
Yes
REST API supported
Yes
Ingest pipelines
No
50GB recommended max
No
1 replica of each shard used in query execution
Intermediate—
Limited to 64GB JVM; forces horizontal scaling
Intermediate—
Requires horizontal scaling
Yes
Horizontal scaling supported
Yes
Elastic Cloud Serverless
Yes
Elastic Cloud Serverless
Yes
Profiles available
Yes
Cache controls supported
No
Not supported
Intermediate—
Only source is compressed
Intermediate—
Approximate only
Yes
Supported
Intermediate—
Supported with limited aggregation types and full scans required to update
Intermediate—
Millions/sec but requires high resource footprint
No
Not supported
No
Limited to JSON/CSV, relies on Logstash.
No
Not supported
No
Not supported
No
Not supported
No
Not supported
Intermediate—
Requires Logstash
Yes
Supported

Feature comparison of ClickHouse and Elastic
Featured
Object storage for reads and writes in OSS	Yes Native object storage support	No Paid, Elastic Cloud Serverless only
Accurate aggregations for high cardinality	Yes Aggregate states for accuracy	No Terms aggregations are estimates
Full columnar storage with high compression	Yes Columnar storage with high compression	Intermediate— Doc values provide columnar storage, not compressed
Streaming ingestion support	Yes ClickPipes for streaming ingest	Intermediate— Requires Logstash/third parties, no hosted ingestion
Native JSON support with type preservation	Yes JSON with type fidelity	Intermediate— First event field determines type
High query concurrency (1,000+ QPS per node)	Yes 1,000+ QPS per node	Intermediate— Requires horizontal scaling/replicas
Parallelized query execution	Yes Full parallelization within & across shards	Intermediate— Limited; relies only on shards
Full join support for correlations	Yes Full join support	Intermediate— LOOKUP JOIN only
Support for roll ups, downsampling and transforms	Yes Materialized views (incremental & refreshable)	Intermediate— Limited types; requires full scans
SQL support	Yes Full SQL support	Intermediate— Limited syntax coverage
Natural language search	Yes Natural language search via HyperDX	Yes Natural language search supported
Flexible deployment (self-managed & cloud)	Yes Self-managed & cloud	Yes Self-managed & Elastic Cloud
Shared-storage architecture with decoupled compute and storage	Yes Decoupled compute/storage (ClickHouse Cloud)	Yes Elastic Cloud Serverless
Stateless compute nodes with fast warm-up time	Yes Stateless compute nodes (ClickHouse Cloud)	Yes Elastic Cloud Serverless
Real-time ingest	Yes Real-time ingest supported	Yes Real-time ingest supported
Micro batch/single row inserts	Yes Async inserts for small batches	Yes Inserts supported
Inverted index support	Yes Inverted index support	Yes Inverted index supported
HTTP REST interface	Yes REST API support	Yes REST API supported
Data transformation at ingest time	Yes Incremental materialized views	Yes Ingest pipelines
Unlimited shard sizes	Yes Unlimited shard sizes (disk-based)	No 50GB recommended max
Distributed query execution across replicas	Yes Parallel replicas distribute queries	No 1 replica of each shard used in query execution
Vertical scaling to support large machines	Yes Vertical scaling to very large machines	Intermediate— Limited to 64GB JVM; forces horizontal scaling
High query concurrency per node	Yes Up to 1,000 concurrent queries per node	Intermediate— Requires horizontal scaling
Horizontal scaling with sharding	Yes Horizontal scaling with sharding	Yes Horizontal scaling supported
Read/write isolation	Yes Read/write isolation (ClickHouse Cloud)	Yes Elastic Cloud Serverless
Elastic scalingUp/down with load	Yes Elastic scaling with load (ClickHouse Cloud)	Yes Elastic Cloud Serverless
Custom hardware profiles in Cloud offering	Yes Custom hardware profiles available	Yes Profiles available
Granular cache control	Yes Granular cache control	Yes Cache controls supported
Join performanceMulti-billion rows	Yes Efficient joins across billions of rows	No Not supported
High data compression	Yes High compression built-in	Intermediate— Only source is compressed
Support for high cardinality aggregations	Yes Accurate aggregations at scale	Intermediate— Approximate only
High concurrency performance	Yes 1,000+ concurrent queries per node	Yes Supported
Support for downsampling and transforms to accelerate queries	Yes Materialized views (incremental & refreshable)	Intermediate— Supported with limited aggregation types and full scans required to update
High insert rates	Yes Tens of millions/second	Intermediate— Millions/sec but requires high resource footprint
File format support for querying in place	Yes 70+ formats including Parquet, ORC, Avro, JSON, CSV	No Not supported
File format support for data load	Yes 70+ formats including Parquet, ORC, Avro, JSON, CSV	No Limited to JSON/CSV, relies on Logstash.
Query external data in-place	Yes Query in-place via table engines (e.g. Postgres, S3)	No Not supported
Hosted Ingest to read from Kafka and S3	Yes Hosted ingest from Kafka & S3	No Not supported
Support for open table and file formatse.g. Iceberg, Parquet, ORC	Yes Open format support	No Not supported
Support for third-party catalogs	Yes Third-party catalog support	No Not supported
External table engines	Yes Connect to Postgres, MongoDB, MySQL, S3, Kafka, and more	Intermediate— Requires Logstash
Support for Open Telemetry	Yes Supported	Yes Supported

Lower costs

10x cost savings thanks to high compression and resource efficiency

Simpler at scale

Homogenous architecture and vertical scaling simplifies and reduces nodes

Built for high cardinality analytics

Column orientation designed for high cardinality queries

Open and interoperable

Supports open standards like OpenTelemetry and integrates directly with systems and formats such as Postgres, Kafka, Parquet, and Iceberg.

Migrate your workload from Elastic today

Cut costs, boost performance, and unlock observability at scale with ClickHouse.
We’ll get you started on a 30 day trial and $300 credits to spend at your own pace.

More comparisons

ClickHouse vs PostgreSQL

ClickHouse vs Redshift

ClickHouse vs Snowflake

Products

Resources

Company

Join our community

Comparisons

Partners

AWS
Azure

Stay informed on feature releases, product roadmap, support, and cloud offerings!

Loading form...

Trademark Privacy Security Legal Cookie policy