Truly's global telephony back end has been battle tested with organizations with hundreds of users over the past 8 years. It has achieve 99.999% uptime in 2021 due to a sophisticated system with heavy redundancy built in. This article discusses this redundancy.
Data Center Footprint
3 AWS regions (Virginia, Frankfurt, Sydney).
Multiple Availability Zones in each region
Load Balancing
Requests globally loadbalanced via Route53 to the nearest available data center
Within each region, requests are loadbalanced to one of several redundant SBCs
Within each AZ, requests are loadbalanced to one of several App/Telephony Servers
Rate limiting at the perimeter at the User/App level, API/IP Adress level and at the SBC unique phone number level.
Database
Core Database RDS (AWS managed), with 5 minute failover to hot standby
Loadbalanced requests across read-write replicas (10:1 read:write provisioning)
Application level redundancy to permit traffic to pass through in cases of partial database failures.
Carrier Redundancy:
At least 3 redundant termination providers in each market, with at last one Tier1 provider