TURN Relay Throughput Benchmark
usrsctp/C DataChannel — 1KB chunks for 10s — same browser, same SCTP, different relay
Server Configuration & Comparison →
Server Comparison Matrix
|
QA KRelay |
Current COTURN |
Experimental COTURN |
STUNner (planned) |
| Software | COTURN | COTURN 4.6.2 | COTURN 4.6.2 | STUNner |
| Language | C | C | C | Go |
| Instances | Managed by DevOps | 3x t3.large | 3x t3.large | TBD |
| Deployment | EC2 ASG + NLB | EC2 ASG + NLB | EC2 ASG + NLB | K8s pods + Gateway API |
| Load Balancer | NLB | NLB (2 AZs) | NLB (2 AZs) | K8s Service |
| Static IPs | EIPs | EIPs (Lambda) | EIPs (Lambda) | Dynamic |
| VPC | Keeper QA | 10.100.0.0/16 | 10.200.0.0/16 | EKS VPC |
| TLS | Yes | TLS 1.3 | TLS 1.3 | TBD |
| Throughput | — | — | — | — |
| Graph | | | | |
1. Direct (STUN only)
No relay — tests if direct UDP works
| Relay | None — direct UDP via STUN hole-punching |
| Note | Fails when either side is behind symmetric NAT |
—
2. QA KRelay — krelay.qa.keepersecurity.com
COTURN — Production QA relay
| Software | COTURN (production QA) |
| Infra | EC2 ASG + NLB (managed by DevOps) |
| Ports | 3478 UDP/TCP, 443/5349 TLS |
| Auth | TURN REST API (shared secret) |
| Network | Keeper QA VPC |
Full comparison →—
3. Current COTURN — krelay.current.keeperpamlab.com
COTURN 4.6.2 — New cluster A
| Software | COTURN 4.6.2 (compiled from source) |
| Instances | 3× t3.large (2 vCPU, 8GB) |
| NLB | 2 AZs, cross-zone, static EIPs |
| VPC | 10.100.0.0/16 (dedicated) |
| Ports | 3478/3479 UDP/TCP, 443/5349 TLS |
| Relay range | 49152–65535 |
| EIP/instance | Yes (Lambda auto-assigns) |
Full comparison →—
4. Experimental COTURN — krelay.experimental.keeperpamlab.com
COTURN 4.6.2 — New cluster B
| Software | COTURN 4.6.2 (compiled from source) |
| Instances | 3× t3.large (2 vCPU, 8GB) |
| NLB | 2 AZs, cross-zone, static EIPs |
| VPC | 10.200.0.0/16 (isolated from current) |
| Ports | 3478/3479 UDP/TCP, 443/5349 TLS |
| Relay range | 49152–65535 |
| EIP/instance | Yes (Lambda auto-assigns) |
Full comparison →—