DB
Back homeIndustry

Capital markets

Collapse the tick store, the in-memory cache, and the stream processor into one deterministic engine.

The problem

Why this is hard today

A trading platform is a chain of systems held together by hope and reconciliation jobs: a tick store, a cache for live state, a stream processor for signals, and a warehouse for research. Every boundary is latency, and every copy is a source of disagreement.

The desks that win optimize the boundary where a strategy reads state it just wrote — and that is exactly the boundary a multi-system stack cannot make consistent.

One engine that keeps fresh writes readable, current state O(1), and counts exact removes the reconciliation problem instead of automating it.

Where NYXDB fits

Use-case journeys

Real-time market data & trading analytics

Fresh-write reads for the live book, exact volumes, and streaming signals — all on one engine.

Market data & trading analytics

Pre-trade & post-trade risk

Read-your-writes per-account state and exact counts behind every risk threshold.

Fraud & risk

Execution observability

Turn raw execution events into live venue/latency metric tables with continuous transforms.

Observability & logs

Architecture

How NYXDB fits capital markets

Ticks append and fan out through PSI to keyed current-state tables and continuous transforms — history, live book, and derived signals in one runtime.

  1. 01

    Tick history

    Append tables hold the full tape, durably journaled.

  2. 02

    Live book

    Keyed tables collapse to the latest quote per instrument.

  3. 03

    Signals

    Continuous transforms maintain OHLC/VWAP-style rollups in place.

  4. 04

    Research

    HISTORICAL and AS OF reads serve backtests off the same data.

Real SQL

Representative query

Exact volume by market
SELECT market, count() AS trades, sum(amount) AS volume
FROM trades
GROUP BY market;

Every statement follows the engine’s own test SQL shapes. See the SQL reference for full syntax.

Capabilities

What you get

O(1) current state

Latest quote per instrument as a point read.

Exact risk math

Deterministic counts and typed aggregates.

Streaming + historical

One table serves the live tape and backtests.

One engine

Retire the cache and the stream processor.

Proof

Measured where it counts

~113×

vectorized aggregate (48.7 → 0.43 ms)

ADR-060

~480–490k

rows/s ingest, end-to-end (local)

ADR-075

µs-classTODO-verify

current-state reads

placeholder — not yet substantiated

Figures marked TODO-verify are placeholders pending a published, reproducible benchmark; substantiated numbers cite their source.

One engine for the trading stack

Explore the market-data solution or run a node.