Separate the price path from the order path
A bridge receives market data and transmits orders, but those flows have different failure modes. We document price sources, aggregation or selection logic, symbol normalization and downstream distribution separately from order routing, execution reports and rejection handling.
Symbol mapping is a production control
Broker symbols and liquidity-provider instruments rarely align perfectly. Mapping must cover names, contract sizes, precision, trading sessions and any instrument-specific rules. A controlled mapping table prevents silent inconsistencies between what clients see and what execution venues accept.
- Price-flow and order-flow architecture
- LP/FIX session and endpoint inventory
- Symbol and contract-specification mapping
- Routing and rejection observability
- Disconnect/reconnect and failover policy
- Bridge acceptance and incident test cases
Routing policy should be observable
Whether routing is simple or conditional, operators need to know which destination was selected and why. We design around explicit routing rules, session state, order lifecycle events, reject reasons and logs so incidents can be investigated without guessing.
Connectivity failure must have defined behaviour
Disconnects, stale prices and unavailable endpoints are normal operational scenarios. Reconnect timing, feed health thresholds, order handling during outages, failover rules and escalation paths should be decided before production rather than during an incident.
Acceptance testing with real failure cases
Testing includes symbol mismatches, rejected orders, partial execution where applicable, stale or missing prices, session resets and endpoint loss. The handover documents what is monitored, who owns each component and which evidence is needed to diagnose an issue.