The Distribution Layer

Gateways

Gateways are the governed layer where applications, agents and external integrations connect to one interface to query data, subscribe to streams and invoke procedures across several Centers.

Function inside the fabric

A single interface in front of many Centers

A Center holds state, and a single Center can serve applications directly. Scale divides it: by asset class, by region, by the boundary between live state and history. Without a layer above, every consumer inherits that division and must be revised each time it moves.

A Gateway answers for the whole estate. Requests are resolved across the Centers behind it, Centers are added and removed without a restart, entitlements and audit apply at the point of entry, and caching, conditional streams and fan-out keep growth in consumers from becoming proportional load on the infrastructure beneath.

Consumers depend on a fixed contract. What sits behind it stays free to change.

A Gateway in front of several Centers, serving client applications, agents and BI tools through one governed interface

Key capabilities

Access, efficiency and control in one layer

Insulating consumers from data estate complexity

  • Aliasing
  • Virtual tables
  • Virtual streams
  • Virtual procedures

Bandwidth and performance optimization

  • Data caching
  • Conditional streams
  • Dynamic subscriptions
  • Fan-out subscriptions

Scalable infrastructure with governance built in

  • Load balancing
  • Dynamic composition and self-healing
  • TLS, with SSO, RBAC and OAuth compatibility
  • Access journal

Example patterns

How Gateways are deployed

Virtual tables over a sharded estate

  • Shards presented as one table, with the Gateway planning the query across them
  • Consumers write the SQL they would write against a single Center
  • Schemas need not match, one shard can gain a column without the others changing
A Gateway presenting several sharded Centers as one logical table to a consumer
Virtual tables. One logical table spanning several Centers and both stores.

Fan-out subscriptions

  • Demand on a trading floor overlaps heavily, the same instruments, adjacent columns, marginally different filters
  • Handled individually, the source bears the full multiple
  • The Gateway consolidates the overlap into a single upstream subscription, then serves each subscriber its own filtered rows and columns
Overlapping consumer subscriptions collapsed into one upstream subscription and fanned back out with per-subscriber filters
Fan-out subscriptions. One subscription upstream, many downstream.

Caching the state of the world

  • Every subscriber needs the current picture, before it can start applying updates
  • A dashboard opened across a floor asks for it again with each session
  • The Gateway caches common queries and the opening snapshot, so the Center only handles work that is new
A Gateway serving cached query results and snapshot state to repeat consumers without reaching the Center
Data caching. Repeated demand answered once.

Gateway to Gateway, across regions

  • Gateways call other Gateways, so a desk connects locally and cross-region requests hop between them
  • No cross-region connection per consumer
  • Each site keeps its own entry point, cache and journal, while the estate behind them stays one logical data set
Gateways in two regions calling one another so consumers reach a local point of entry
Gateway to Gateway. A local point of entry in each region.

Application Delivery

Gateways distribute the data. Web Servers turn it into an application.

Tailored, real-time applications assembled from grids, pivots, charts and trees, running on the data the fabric already holds.

Explore Web Servers Explore the Application Fabric