ResourcesArticles

Building an Anycast Architecture for Programmatic Call Control

Alex Lee, a Telnyx python developer, describes how the product team relied on the strengths of the Telnyx global IP backbone when building our new Call Control product.
At Telnyx, we support applications on the cutting edge of internet communications. By building our products on top of a global network and a telephony engine purpose-built for internet communications, we’re able to maintain top-tier quality, security and service.
When we began scoping out our Call Control product, it was a given that we’d leverage our global infrastructure to maintain low latency API commands and unassailable security—even at a massive scale. But, we also wanted Telnyx Call Control to have as straightforward an implementation as possible for our customers. Achieving both goals required a new anycast architecture modeled off our existing call routing engine.

High Availability, Low Latency Commands and Webhooks

At minimum, a programmatically controlled call consists of two API commands, two responses and two webhooks. The more control functionality introduced, the more times our system and yours need to communicate. A robust implementation might average five to 10 commands per call. When you’re dealing with thousands of calls a day, the number of commands and their responses starts to add up.
Given the volume of Telnyx-user interactions, it’s important the system be designed to always assure commands are received and webhooks are delivered, and the round-trip time between these two actions is minimized. The more we reduce latency and assure availability, the smoother our clients’ applications operate and the happier their users are. Thankfully, we can boast one of the highest performing APIs.
We did this by deploying infrastructure in each of our points-of-presence around the world and building an anycast call control architecture parallel to our existing SIP architecture. Just like our SIP connections, you send all commands to a single endpoint. Using anycast, our system determines the most efficient and available PoP to receive and process the command in the fastest way possible.


Benefits of anycast architecture:
  • A single endpoint for all communications, allowing a more straightforward implementation.
  • Minimized contact with the public internet for better performance and security.
  • Automated failovers and rerouting in the case of a downed proxy.

Enhanced Security with Fully Encrypted Payloads

Our private network minimizes contact with the public internet, but any command will still be vulnerable during those short hops between your network and ours. Considering a heavy user of call control could be sending 50 million commands per month, that’s a lot of proprietary information vulnerable to attack.
We implemented comprehensive encryption for every command and webhook to protect that data no matter whose network it's crossing. By encrypting both routing information and command details, our call control proxy servers and the client’s are the only systems that can determine where a command is headed and what actions, callers, recordings and networks are involved.

By using the site, you agree to our use of cookies. Accept and close Find out more here.