Role in ClickHouse's comparison material

ClickHouse describes a long-standing split in which Elasticsearch was used for text search and ClickHouse for analytics. The post "Do you still need Elasticsearch for log analytics?" argues that this division no longer holds, because ClickHouse now includes a redesigned full-text search powered by inverted indexes, with matching documents fed directly into its vectorized analytical engine. The framing is that logs are analytical data that happen to contain text, so search is usually only the first step before counting, grouping by service, narrowing a time range or isolating hosts.

  • Positioned as the traditional engine for text search
  • Compared on log analytics rather than general search
  • ClickHouse's argument rests on combining search with aggregation in one engine

Source: Do you still need Elasticsearch for log analytics? ClickHouse says no. | ClickHouse

OpenTelemetry log analytics benchmark

The benchmark compares the open-source versions of ClickHouse and Elasticsearch on an OpenTelemetry log workload generated by ClickHouse's OpenTelemetry demo application. Datasets scale up to 50 billion rows, with source data published as Parquet files in a public Amazon S3 bucket in the eu-west-3 region, each file holding one billion log rows sorted by timestamp. ClickHouse reports that it ran the full-text analytical workload 2–6x faster than Elasticsearch across all tested dataset scales, and that it stored the OTel log dataset far more compactly. The post states that the benchmark is reproducible end to end, with source data, schemas, queries, ingestion scripts and measurements published in a public GitHub repository.

  • Datasets up to 50 billion rows
  • Public bucket: s3://public-pme/text_bench
  • Nine queries covering drill-down, error counts, service breakdowns and time-based trends
  • Queries expressed in ClickHouse SQL, Elasticsearch Query DSL and ES|QL
  • Reported result: ClickHouse 2–6x faster on the full-text analytical workload

Source: Do you still need Elasticsearch for log analytics? ClickHouse says no. | ClickHouse

Elasticsearch index configuration used

The benchmark documents how the OpenTelemetry log data was mapped in Elasticsearch so that the comparison reflects a realistic observability schema. Log bodies were made full-text searchable with the standard analyzer, while structured dimensions and map-like attributes used keyword and flattened types respectively. Compression settings were tuned to reduce the on-disk footprint before storage figures were compared.

  • Body mapped as text with the standard analyzer
  • Structured string dimensions mapped as keyword
  • Map-like attribute fields mapped as flattened
  • TraceFlags and SeverityNumber mapped as byte
  • codec: best_compression enabled

Source: Do you still need Elasticsearch for log analytics? ClickHouse says no. | ClickHouse

Migration context with ClickStack

Elasticsearch also appears in ClickHouse's community material as an existing destination in telemetry pipelines. A post written with the Bindplane team describes teams already shipping logs to Elasticsearch, metrics to Datadog and traces to Splunk, and walks through adding ClickStack as an additional destination without a rip-and-replace migration. Bindplane routing logic controls what data goes where, allowing services to be migrated one at a time while the existing backend continues to run.

  • ClickStack added alongside existing backends as an extra destination
  • Routing rules allow per-service migration
  • ClickStack consolidates logs, metrics and traces on ClickHouse

Source: Faster OpenTelemetry migrations to ClickStack with Bindplane | ClickHouse

ClickHouse full-text search capabilities referenced

The comparison with Elasticsearch rests on ClickHouse's text index work. ClickHouse engineers describe redesigning the full-text index to work efficiently on object storage, favouring sequential access patterns so that many queries can be answered directly from the index without reading the indexed text column. The text index is a data skipping index defined per column and stored inside each data part alongside the column files and the primary index. ClickHouse best-practice material lists the text index as a full inverted index for text that reached general availability in version 26.2.

  • Text index redesigned for object storage latency characteristics
  • Stored per data part as a data skipping index
  • Posting lists use roaring bitmap, VarInt-encoded and embedded layouts
  • Text index GA from version 26.2

Sources: Building high-performance full-text search for object storage | ClickHouse, Top 10 best practices tips for ClickHouse | ClickHouse

Sources

  1. Top 10 best practices tips for ClickHouse | ClickHouse https://clickhouse.com/blog/10-best-practice-tips Verified 14 Nov 2026
  2. May 2025 Newsletter | ClickHouse https://clickhouse.com/blog/202505-newsletter Verified 14 Nov 2026
  3. June 2025 Newsletter | ClickHouse https://clickhouse.com/blog/202506-newsletter Verified 14 Nov 2026
  4. Faster OpenTelemetry migrations to ClickStack with Bindplane | ClickHouse https://clickhouse.com/blog/bindplane-faster-otel-migrations-to-clickstack Verified 14 Nov 2026
  5. Building high-performance full-text search for object storage | ClickHouse https://clickhouse.com/blog/clickhouse-full-text-search-object-storage Verified 14 Nov 2026
  6. Do you still need Elasticsearch for log analytics? ClickHouse says no. | ClickHouse https://clickhouse.com/blog/elasticsearch-log-analytics-clickhouse Verified 14 Nov 2026

Last verified 14 Nov 2026. This entry is compiled from the public web pages listed above. Nothing here is stated that those pages do not, and each of them was read on the date shown.