Geographic locations and the latency.

We are recording redistributing the data across all the sources simultaneously in this data centers:

 * AWS US-EAST-1

 * AWS EU-CENTRAL-1

 * EQUINIX WA1

Historical data delivered using the REST API (all /history and several /latest endpoints) and the flat files are generated in the separate process where we replay all the markets (seen from different locations) and create single data feed prone to issues related to specific geographic areas like network issues etc.

Real-time data delivered using the REST (all /current and several /latest endpoints), WebSocket and the FIX API originates from the specific site where the customer was connected. On the Enterprise plan, it's possible to have direct access to the sites or setup AWS VPC Peering in that case latency between our infrastructure and the customer is below one millisecond. If the connection was established to the public endpoint, then it's routed using the GeoDNS to the closest site; latency then depends on the distance between the customer and the site where the connection was routed [~20ms].