Learn about latency and server location

AI-generated Key Takeaways

• Place your servers near relevant Google Cloud regions and configure DNS to route traffic efficiently to meet latency requirements.

• Using an HTTP persistent connection is strongly recommended to reduce latency by reusing connections for segment requests.

• While requests are often sent from the closest region, placing servers near multiple key regions helps receive a more complete set of requests for your target locations.

• The deadline for sending a segment response is 50 ms, including network time and server processing.

To help meet the latency restrictions of the Real-time Curation service, place your servers near the Google Cloud regions most likely to send segment requests to locations you're targeting. Provision and configure DNS such that traffic sent to your endpoint from a given region is routed to your closest server.

The following table lists Google Cloud regions used in Real-time Curation, their approximate location, and examples of targeted locations that are most likely to be sent to your servers in a SegmentRequest from those regions:

Google Cloud Region	Region Location	Example target location
us-east1	South Carolina, United States	North America (East Coast)
us-west1	Oregon, United States	North America (West Coast)
us-central1	Iowa, United States	North America (Central)
europe-west1	Belgium	Europe
europe-west4	Amsterdam, Netherlands	Europe
asia-southeast1	Singapore	Asia
asia-east1	Taiwan	Asia

Tip: Depending on the locations you're targeting, the optimal region to place your servers near might not be obvious. For example, Real-time Curation doesn't send segment requests from a region that is directly located in South America, but the majority of segment requests from users in that location are typically sent from any of us-east1, us-west1, or us-central1 regions. If you're unsure which regions are best suited for a location you're targeting, discuss it with your account manager.

Use an HTTP persistent connection to reduce latency

To reduce latency, Google strongly recommends that you configure your Real-time Curation integration to use a persistent connection. After a persistent connection is established, your endpoint will reuse connections rather than create a new one for each incoming segment request.

Server location

While you can expect that most segment requests are sent from the region that is closest to the user's location, Google doesn't guarantee that to always be the case. As you place servers close to a greater number of regions, you can expect to receive a more complete set of segment requests for your targeted locations. Google recommends placing servers near a subset of regions appropriate for your targeted locations. For example, most North American traffic can be received by placing your servers close to the us-east1, us-west1, and us-central1 regions.

The deadline that you must send a segment response by is 50 ms.

This deadline includes both the network time between the region and your server, and the time for your server to build a response. We recommend you have at least a 10 ms buffer for unexpected changes in network latency.

Next Steps

• Real-time Curation protocol reference: Learn how the requests and responses used in Real-time Curation are structured.
• Parse Real-time Curation requests: Learn how to parse Real-time Curation requests.
• Build Real-time Curation responses: Learn how to build Real-time Curation responses.