One Live Feed for Every Trading Tool
Capital MarketsFeeding every internal tool from one upstream connection instead of a new vendor socket per feature, scaling to hundreds of thousands of ticks a second.
Overview
A registered investment adviser and a separate quantitative research operation needed every internal tool fed by the same live market data, without paying for a new vendor socket per feature or capping throughput as volume grew from thousands to hundreds of thousands of ticks a second. Canvas Labs Partners built two fan-out layers from a single upstream connection: one tuned for sub-100ms latency, one for order-execution throughput.
7,000+ ticks/sec at under 100ms
Redis fan-out, running for under $150 a month
100,000s of ticks/sec
Kafka backbone, running for under $100 a month
One feed serves every app
a single upstream connection, not one socket per tool
Challenge
Two different trading operations, a registered investment adviser and a separate quantitative research desk, needed every internal tool fed by the same live market data. Paying for a new vendor socket per feature was not sustainable, and neither was a pipeline that capped out as usage grew from thousands of ticks a second to hundreds of thousands.
Solution
For the registered investment adviser, Canvas Labs Partners decoupled the upstream market-data provider sockets from every downstream consumer using an in-memory Redis fan-out, tuned to hold latency under 100 milliseconds end to end regardless of how many tools were reading the same feed. A single provider connection now serves every internal tool the adviser runs.
The quantitative research operation needed more raw throughput than the Redis layer was built for, so Canvas Labs Partners built a separate, higher-throughput order-execution backbone on Apache Kafka. Where the Redis fan-out holds latency under 100 milliseconds, the Kafka backbone carries volume into the hundreds of thousands of ticks per second the quant desk's strategies generate.
Both systems share the same underlying principle: one upstream connection, fanned out with WebSockets and pub/sub to however many internal consumers need it, rather than a new vendor socket bought and billed for every feature. That structure is what kept the running cost of each implementation low while throughput scaled by orders of magnitude between the two operations.
Results & benefits
The Redis fan-out for the registered investment adviser now carries more than 7,000 ticks a second at under 100 milliseconds of latency, running for less than $150 a month. The Kafka backbone built for the quantitative research operation carries volume in the hundreds of thousands of ticks per second, for under $100 a month. Both systems reduce to the same shape: one upstream connection serving every consumer, rather than a new vendor socket opened and paid for per feature, so adding another internal tool to either operation no longer means a new line item on the data bill.
Customer
Registered investment adviser & quant desk
Services
- Trading & quantitative infrastructure engineering
- Real-time data streaming architecture
- Cost-efficient infrastructure design
Tech stack
About us
Canvas Labs Partners is a software consulting team that designs and ships production agentic AI systems: multi-agent pipelines, tool-calling architectures, and human-governed automation, together with the trading infrastructure, blockchain, mobile, and product engineering those agents plug into. We publish our work as a case ledger rather than a portfolio, so every claim on this site links to the engagement behind it. Engagements follow the same four-step discipline, Map, Bound, Instrument and Govern, whatever the stack, and every entry in the ledger carries a result we are willing to be measured against.
Delivered individually, in a prior role, by one of Canvas Labs Partners’ founding team members. Not by the firm itself.
Have a system like this one?
We built this one end to end. If you are facing something similar in capital markets, book a free discovery call and we will tell you what it would take.