Kdb+ Alternatives for Time-Series and Tick Data

Kdb+ is KX's columnar time-series database and the q language that runs it, used for decades on trading desks for tick data, risk and research. It is alive and has a new wrapper. KX announced on November 17, 2025 that KDB-X is generally available, describing it as built on the same kdb+ core with existing code and data still compatible, plus vector and AI data handling. The KX pricing page offers KDB-X free to start, with usage-based pricing sized by memory (GiB) for larger setups. KX publishes no rates, and custom deals go through sales.

The free tier has hard limits. The KDB-X Community Edition terms allow personal and commercial projects but cap q at 16 GB of RAM, 24 CPU cores, four secondary threads and 16 connections per process, on one instance. Teams outgrow that quickly, and q is a scarce skill to hire for. The seven databases below cover tick storage, SQL analytics, embedded research and streaming metrics, and most of them have a fully open source edition.

Prices below are as of November 8, 2026. Our method is on How we review tools.

Quick comparison

QuestDB

QuestDB maps most directly onto kdb+ habits. The QuestDB repository lists SQL extended for time series with SAMPLE BY, LATEST ON, ASOF JOIN and WINDOW JOIN, plus materialized views and n-dimensional arrays. Data arrives over its own binary protocol, InfluxDB Line Protocol, the PostgreSQL wire protocol or a REST API with CSV import.

The core is Apache 2.0 and has about 17,300 GitHub stars. QuestDB's pricing page says Enterprise is priced by sales and adds replication with failover, role-based access down to columns, TLS, SSO, point-in-time recovery, and Parquet and Iceberg tables on object storage. It can run self-managed or in your own AWS or Azure account.

Best for: Market data teams who want as-of joins and bar sampling in SQL instead of q.

Pricing:

• Open Source: free, Apache 2.0
• Enterprise: contact sales; free trial
• BYOC on AWS or Azure: through Enterprise

Watch out for:

• The open source edition has only basic authentication over unencrypted connections. TLS and access control are Enterprise features.
• High availability and replication are also Enterprise-only, so a production cluster means a sales conversation, much like KX.

Official site: QuestDB.

ClickHouse

ClickHouse is a column store built for scanning huge tables fast, and many firms run it next to or instead of kdb+ for historical analysis. It speaks SQL, supports ASOF JOIN for matching trades to quotes, and scales across shards and replicas. The server is open source under Apache 2.0, so you can self-host it on any number of cores without a licence.

The ClickHouse Cloud pricing page lists storage at $25.30 per TB a month and compute from $0.2181 per compute unit-hour on Basic, $0.2985 on Scale and $0.3903 on Enterprise, for AWS us-east-1 in US dollars. New accounts get $300 of credit for 30 days.

Best for: Analytics teams that want SQL at kdb+ speeds over very large history.

Pricing:

• Open source: free, Apache 2.0
• Cloud Basic: $0.2181 per compute unit-hour, up to 1 TB storage
• Cloud Scale: $0.2985 per compute unit-hour
• Cloud Enterprise: $0.3903 per compute unit-hour
• Storage: $25.30 per TB a month; trial $300 credit for 30 days

Watch out for:

• It is tuned for batch inserts. Row-by-row updates and deletes are slow mutations, so a live order book needs a different design.
• Cloud rates vary by provider and region, so the us-east-1 figures are only a starting point.

Official site: ClickHouse.

DolphinDB

Of all the picks, DolphinDB works most like kdb+. It is a distributed time-series database with a full scripting language, a stream-processing framework and built-in analytics functions, so research and production can run in one system. Plugins connect it to market data feeds, and a November 6, 2026 news post announced one for XTP-Pro.

The DolphinDB site also lists Swordfish, an embedded low-latency engine, Shark for CPU and GPU computing, and Orca for enterprise real-time work. Pricing is not published. You request a trial or contact sales, and the docs say memory, cores and cluster nodes are constrained by your licence.

Best for: Quant teams who want kdb+'s all-in-one model with a different vendor.

Pricing:

• Trial: on request
• Licences: price on request; memory, cores and nodes set by licence

Watch out for:

• You swap q for another proprietary language, so the hiring problem changes shape rather than going away.
• Like KX, it has no public price, which makes cost comparisons slow.

Official site: DolphinDB.

TimescaleDB

TimescaleDB turns PostgreSQL into a time-series database with hypertables that partition by time, columnar compression and continuous aggregates. Its company now trades as Tiger Data. The appeal for kdb+ teams is ordinary PostgreSQL: any driver, BI tool or ORM works, and SQL skills are easy to hire.

The Tiger Data pricing page lists Tiger Cloud Performance from $30 a month and Scale from $36 a month for compute, with storage at $0.177 and $0.212 per GB-month. Usage is metered hourly and billed monthly in arrears. There is a 30-day trial with no card, and self-managed TimescaleDB is free.

Best for: Teams that want time-series features without leaving PostgreSQL.

Pricing:

• Self-managed: free
• Tiger Cloud Performance: from $30 a month compute, $0.177 per GB-month storage
• Tiger Cloud Scale: from $36 a month compute, $0.212 per GB-month storage
• Enterprise: contact sales

Watch out for:

• The repository mixes Apache 2.0 code with a separate Timescale License directory, so check which features you may offer as a hosted service.
• Row-store PostgreSQL underneath means wide scans run slower than in kdb+ or ClickHouse until data is compressed.

Official site: TimescaleDB.

DuckDB

For quants who used kdb+ mainly as a fast research engine on one machine, DuckDB is the free answer. It runs inside Python, R or a command line, reads Parquet and CSV files in place, and supports ASOF JOIN and window functions in standard SQL. Nothing needs a server.

Version 2.0.0 reached PyPI on October 21, 2026, and the whole engine is MIT licensed. MotherDuck sells a hosted service built on it for teams that want shared storage.

Best for: Researchers who query Parquet tick files on a laptop or a single large server.

Pricing:

• Free, MIT licence
• MotherDuck hosted service: priced separately

Watch out for:

• Only one process can write to a database file at a time, and other processes can only read it, per DuckDB's concurrency docs.
• There is no streaming ingest or tickerplant equivalent, so live capture has to happen elsewhere.

Official site: DuckDB.

InfluxDB 3

InfluxDB 3, from InfluxData, was rebuilt on Apache Arrow, DataFusion and Parquet, and it answers SQL as well as InfluxQL. It suits the monitoring and sensor side of a firm more than order books: metrics come in through line protocol, the same format QuestDB accepts, and Telegraf agents can feed it.

The InfluxData pricing page lists InfluxDB 3 Core as free and open source. Enterprise adds high availability, read replicas, clustering and SSO, is priced by CPU configuration, and has a 30-day trial. Cloud Serverless charges $0.0025 per MB written and $0.002 per GB-hour of storage.

Best for: Infrastructure and IoT teams collecting high-frequency metrics.

Pricing:

• Core: free, open source
• Enterprise: custom price by CPU configuration; 30-day trial
• Cloud Serverless: $0.0025 per MB written, $0.012 per 100 queries, $0.002 per GB-hour storage
• Cloud Dedicated: custom

Watch out for:

• Core is single-node with no high availability, so production redundancy needs Enterprise.
• Its query model fits metrics. Complex as-of joins across trade and quote tables are not its strength.

Official site: InfluxDB 3.

Apache Druid

Apache Druid targets a narrower job: sub-second queries on event streams for live dashboards. It ingests straight from Kafka or Kinesis, indexes data as it lands, and serves many concurrent users. A trading firm might use it for a real-time P&L or surveillance screen rather than for research.

The Druid downloads page lists version 38.0.0, released October 1, 2026. It is an Apache Software Foundation project under the Apache 2.0 licence, with commercial hosting from vendors such as Imply.

Best for: Teams building high-concurrency dashboards on streaming data.

Pricing:

• Free, Apache 2.0
• Managed hosting: priced by third-party vendors

Watch out for:

• A cluster has several process types, including brokers, historicals and coordinators, so operations are heavier than a single kdb+ process.
• Joins are limited compared with SQL warehouses, so tick-by-tick as-of logic is a poor fit.

Official site: Apache Druid.

How to choose

When to stay with Kdb+

Stay with kdb+ if your desk already runs a tickerplant, real-time and historical databases and a library of q code. Rewriting that is a long project, and KDB-X keeps existing code and data compatible while adding vector search. Individuals and small projects can also test their data against the 16 GB cap of the free KDB-X tier before moving.

Frequently asked questions