berita perusahaan tentang SereneDB launches superfast S3 data search engine.

German startup SereneDB’s Krummelanke is claimed to be the fastest database search engine in the world, combining search and analytics in one product.

The company states AI agents fire hundreds of hybrid find-and-count queries per second, yet every database in production was built for human users. Its Krummelanke engine is designed to handle high-speed agent-scale database search. Krummelanke is open-source under Apache 2.0, compatible with Postgres and Elastic, and can run full-text search over data stored on S3.

Co-founder and CEO Alexander Malandin said: “The agentic future is, above all, a data-load problem. Today an AI agent is a single bee visiting a single flower. What is coming is the swarm: more agents than employees in every organization, each firing hundreds of queries a second. Infrastructure built for human users is neither fast enough nor cost-efficient enough to survive that load. The agentic future requires a new generation of database: ultra-fast search and ultra-fast analytics in one engine, where search operations are measured in nanoseconds and a question over billions of records comes back in milliseconds. One that treats agents as first-class users, withstands their scale, and reaches data wherever it lives. That is what we built, and with SereneDB Krummelanke we are putting it into production.”

SereneDB notes analytical engines like Snowflake, BigQuery, DuckDB over Parquet, and lakehouses generally read columnar data straight from S3 by default. Large sequential scans fit the latency profile of object storage, unlike full text search. Full text search requires an inverted index that lists content items such as words or numbers and maps them to their locations within tables or documents. Inverted index search generates many random accesses. Typically, Lucene-based search engines store indexes on fast local storage separate from the main lakehouse data store. SereneDB says “searchable” data within an Elastic deployment means maintaining a second copy on flash, the most expensive storage tier available.

We asked how SereneDB’s search differs from general object storage data search such as Cloudian HyperStore search.

CTO Andrey Abramov told us: “Cloudian is an S3-compatible storage vendor and HyperSearch is a service attached to the bucket. It collects metadata of stored objects and indexes that metadata inside an underlying search component; this is a separate indexing system populated via data copying. Many services implement search this way and they work well.”

At the search engine core level, however: “There are only a handful of core search technologies and libraries worldwide: Lucene, Tantivy, and IResearch built by SereneDB. Elasticsearch, OpenSearch, Solr and many others are Lucene derivatives and follow a copy-first model. They cannot search data they do not hold locally, so any data to be searched must be extracted, transferred and duplicated into the search cluster.”

SereneDB works differently: “SereneDB indexes data where it already resides. Parquet files and Iceberg tables on S3 remain the source of truth. We store only the derived index. The search itself — full-text with BM25 ranking, vector and geo search — runs against that index with a small local hot set. The Iceberg use case serves as a concrete example: a customer writes changes to an Iceberg table on S3, SereneDB picks up those updates, and serves search and analytics over them. You may point any number of SereneDB nodes at the same table.”

We asked whether object storage, database or lakehouse vendors could adopt SereneDB’s search engine as an alternative to Lucene and its derivatives. Abramov said: “Yes, absolutely! We can not only replace ElasticSearch, OpenSearch, Solr, CrateDB which are pure Lucene derivatives, but also augment them with analytical capabilities. For example, we maintain dedicated documentation covering migration from ElasticSearch.”

The firm has run benchmarks comparing its search engine speed against popular alternatives and outperforming them. Abramov said: “Our goal with these benchmarks was to demonstrate that SereneDB is a faster and modern alternative to the traditional search stack.”

Benchmark Details

SereneDB’s SearchBench includes 92 search and analytics queries executed against 100 million and 1 billion OpenTelemetry log records, the TextBench dataset published by ClickHouse with 15 stored columns per record. All tests run on a single server. Every competing engine was tuned using published configuration parameters, with a 60-second timeout per query. Any query exceeding this limit is marked as did not finish.

Benchmark results for more than a dozen popular search engines are available on a dashboard. Tested engines include Elasticsearch across four configurations, OpenSearch, ParadeDB, CrateDB, ArangoDB, TigerDB with row and column store variants, and vanilla Postgres, with additional engines to be added. Timing uses three runs per query, taking the best result from the final two runs. One query is executed at a time from a single client and result caching is disabled. The dashboard measures latency rather than throughput.

The dashboard’s result cells show performance ratios relative to the fastest engine for each query, where the fastest result equals 1.0x. The overall score is calculated as the geometric mean of these 92 per-query ratios. SereneDB explains this standard ratio averaging method prevents a single outlier query from skewing results.

SereneDB achieves a geometric mean of 1.05x on the 100 million record dataset and 1.07x on the 1 billion record dataset. Across all 92 queries, it is either the fastest engine or within a few percent of the top speed. ElasticSearch ranks second with a 2.8x geometric mean on the 100 million dataset. TigerData column store scores 2.24x on the 1 billion dataset, with numerous timeouts. Vanilla Postgres reaches 204x on the 100 million dataset and 50.69x on the 1 billion dataset.

To simplify comparison across vendors beyond the complex dashboard view, we created a chart showing ordered geometric mean scores from the 100 million benchmark based on screenshot data.

Performance gaps are extremely wide, so we excluded scores above 10 and redrew the chart. As SereneDB states, its search engine leads the tested group.

SereneDB founding and funding

SereneDB was founded in Berlin in 2025 by Alexander Malandin, CEO with prior enterprise sales experience at EMC, Dell and ArangoDB; Andrey Abramov, CTO with search R&D background at Quest Software, EMC and ArangoDB; and Valery Mironov, core engineer previously at ArangoDB and YDB. The eight-person engineering team has collaborated for more than 16 years. Its search core IResearch has been under development since 2014.

The startup announced $2.1 million in pre-seed funding in December 2025, led by Entourage and High-Tech Gründerfonds.

SereneDB Krummelanke is open source under the Apache 2.0 license and available on GitHub.

Beijing Qianxing Jietong Technology Co., Ltd.
Sandy Yang/Global Strategy Director
WhatsApp / WeChat: +86 13426366826
Email: yangyd@qianxingdata.com
Website: www.qianxingdata.com/www.storagesserver.com
Business Focus:
ICT Product Distribution/System Integration & Services/Infrastructure Solutions
With 20+ years of IT distribution experience, we partner with leading global brands to deliver reliable products and professional services.
“Using Technology to Build an Intelligent World”Your Trusted ICT Product Service Provider!