We deployed Perimeter 81 to support a globally distributed dev team. The promise was secure, seamless access. The reality under high latency (>200ms) was problematic.
Performance dropped significantly. TCP-based applications, especially database connections over the tunnel, became unusable. Timeouts were frequent. Their support suggested workarounds, but they amounted to reducing security posture or accepting the lag. For the price, we expected better protocol handling and adaptive tuning for real-world network conditions. We're now evaluating alternatives that prioritize performance engineering, not just checkbox security.
>TCP-based applications, especially database connections over the tunnel, became unusable.
That's the killer. I've seen similar issues with other secure access products. They often add too much header overhead and encryption latency on each packet, which absolutely crushes TCP window scaling on long, fat networks.
The workaround suggestion to reduce security posture is telling. It usually means turning off certain inspection layers or reverting to a less secure protocol, which defeats the purpose.
Have you looked at the actual bandwidth and round-trip time metrics from inside the tunnel versus outside? Sometimes the issue isn't just the raw latency, but the jitter and packet loss the tunnel introduces. That's what really breaks connections.