KX: time-series infrastructure, KDB-X and licence boundaries

KX supplies database and analytical technology for demanding data workloads, including financial markets. KDB-X combines time-series, streaming and vector-oriented processing around the kdb+ environment. For a brokerage, the relevant applications include analysing market data, reconstructing activity and supporting research or operational analytics. The institution builds its application and control logic around that infrastructure.[1]

The useful design question is how current events, historical records and analytical outputs remain consistent. A fast query over incomplete or incorrectly labelled data can still give the wrong business answer. Data definitions, identifiers, timestamp handling and reconciliation therefore belong to the application design, alongside the choice of database.

The current KDB-X installation guide supports Linux on x86_64 and ARM, macOS on Intel and Apple Silicon, and Windows 10/11 through Windows Subsystem for Linux. It explicitly says native Windows installation is not currently supported. That installation scope should be kept separate from Windows-related fixes appearing in release notes for the underlying runtime.[2][8]

The Community Edition licence permits personal and commercial projects without a licence fee, within published resource and deployment restrictions. This is a proprietary licensed route; access to a free edition does not provide unrestricted deployment rights.[3]

The public pages contain a material discrepancy: the product comparison lists 8 connections and 16 GiB RAM, while the detailed licensing page states 16 connections and 16GB. A separate pricing slider advertises free use up to 32 GiB. Those descriptions should be resolved against the licence actually issued before sizing a deployment; the largest marketing figure is not used here as an entitlement.[1][3]

Commercial fees and permitted use sit in the order form. The published standard terms grant an internal-use licence and restrict making software functionality available to third parties. A broker procuring internal analytics and one offering database functionality to customers therefore need different contractual questions answered. The reviewed documents do not establish one universal commercial price.[4]

Both KDB-X editions need a valid licence. The installation guide describes online and air-gapped routes, with a Community licence obtained through the Developer Center and a commercial licence supplied separately by KX. Product installation and licence configuration are therefore part of the same deployment process.[2]

The DB Service quick-start documents a single-node Docker Compose route and a clustered Kubernetes route, with REST, q and Python clients. It also states that Community licences from before May 2026 need to be generated again for DB Service use. Version, licence generation and deployment mode all matter when diagnosing a startup problem.[5]

Operational acceptance should exercise both normal processing and recovery. For example, rebuilding a time-series dataset requires the restored history, later events and analytical queries to agree on which records are present.

The KDB-X guide describes typical enterprise response targets of one hour for Severity 1 on a 24/7 basis, four business hours for Severity 2 and one business day for Severity 3. The customer’s licence and support agreement determine the applicable commitment. These are response descriptions, not guaranteed restoration times. Community channels do not carry the same guaranteed support.[6]

A useful incident record identifies the product version, operating system and architecture, steps to reproduce, error output and business impact. Include the relevant licence and a suitably anonymised example where needed; otherwise an entitlement limit can be mistaken for an application defect.[6]

KX is supplying software. Market-data usage rights, permission to redistribute data and authority to offer a regulated financial service have to be established separately from the database licence. KX’s standard terms place responsibility on the customer for rights to its input data, testing before production, backups and monitoring.[4]

The website terms identify KX Systems, Inc. as the site owner. Product contracting depends on the actual software agreement and order form, so website ownership is a company reference rather than a substitute for identifying the supplier to a particular customer.[7][4]

The release record dates KDB-X general availability to 17 November 2025. Versioned releases and subsequent fixes matter to implementation: a feature in current documentation should be matched to the installed version rather than assumed to exist in every earlier deployment.[8]

Operator

KX Systems, Inc.; the customer agreement identifies the contracting supplier

Service scope

Time-series, streaming and vector data infrastructure, with KDB-X, DB Service and KDB.AI used to develop analytical applications.

Customer types

Financial institutions; Trading and research teams; Application developers

Delivery

On premises; API

Pricing publication

Public and quoted pricing

Platforms

KDB-X; KDB-X DB Service; KDB.AI

Pricing scope

Free Community Edition within its licence restrictions and commercial arrangements. Conflicting public RAM and connection descriptions are detailed in Pricing & commercial terms.

Implementation

Licensed installation with online and air-gapped routes; DB Service supports container and Kubernetes configurations. Application development, data modelling and recovery design remain part of the deployment.

Support

Commercial KDB-X clients use the support portal or [email protected]. Community Edition uses community channels; KDB.AI lists [email protected]. Contract-specific response commitments apply.

Permissions and regulatory scope

Technology supply and software licensing. No financial-services permission inventory is asserted; market-data rights and the customer’s financial permissions are separate.

Provider metadata checked 20 Oct 2026Source 1Source 2Source 5Source 6Source 7Source 4Source 9Source 3

KDB.AI can retain vectors and references while an external kdb+ database holds the full records. That avoids duplicating the complete payload, but new records still require vector ingestion and index synchronisation. The architecture divides the analytical work between the external database and the vector layer; it does not make the index self-maintaining.[9]

KDB.AI’s troubleshooting documentation also flags a practical diagnostic trap: reaching the Community licence’s qIPC connection limit can appear as an authentication or access error. This is a reason to check capacity and entitlement alongside credentials before changing the application’s security configuration.[9]