Question

How to avoid TLS handshake latency on Profiles API?

Product

Twilio Segment

Environment

Segment Console

Answer

Segment's Profiles API has a connection idle timeout of 60 seconds and utilizes TLS 1.2. If a connection is dropped due to inactivity, a full TLS handshake is required to re-establish it. This handshake can add 400ms+ of latency to the first request sent after a connection has timed out.

To avoid this overhead, we recommend one of the following optimization strategies:

Option 1: Adjust Client-Side Keep-Alive Timeout (~30s-40s)

If your application’s keep-alive timeout is higher than Segment's (60s), our load balancer will terminate the connection first. This often results in your client attempting to use a "dead" connection, leading to a retry or a forced handshake.

• Action: Set your client-side keep-alive timeout to a value lower than ours (we recommend starting around 30s-40s).

• Result: This allows your server to gracefully close and recycle the connection on its own terms, preventing unexpected termination by the Segment load balancer.

Option 2: Implement a "Heartbeat" or Proactive Polling (every 20s-30s)

For use cases requiring the absolute lowest possible latency for every request, you can prevent the connection from ever idling out.

• Action: Send a "dummy" or lightweight request (a heartbeat) every 20–30 seconds.

• Result: This keeps the TCP/TLS connection "warm" and active, ensuring the pipe stays open so that production requests never encounter the 400ms handshake penalty.