I've been evaluating Banyan Security for our global team, and while the Zero Trust model is compelling, I'm concerned about the performance overhead for our APAC-based developers. Our primary workloads are SSH and HTTPS connections to on-premises development environments in the US West Coast.
I ran a series of controlled latency tests, comparing direct connection (via a traditional VPN concentrator) against Banyan's service. The setup:
* **Tool:** `tcpping` (measuring TCP handshake + ACK times)
* **Target:** A test server in us-west-2 (AWS)
* **Origins:** Company offices in Singapore (SIN), Tokyo (NRT), and Sydney (SYD)
* **Protocol:** HTTPS/443
* **Sample:** 1000 sequential requests, 1-second interval
**Raw Results (Average Latency in ms):**
| Origin | Direct to Target | Via Banyan (closest PoP) | Delta |
|--------|------------------|--------------------------|-------|
| SIN | 172 | 189 | +17 |
| NRT | 112 | 135 | +23 |
| SYD | 162 | 195 | +33 |
The delta is consistent and appears to be the cost of the additional TLS termination and policy enforcement at the Banyan Access Tier. For interactive SSH sessions, an extra 20-30ms is perceptible but often acceptable. For frequent, small HTTP API calls (like those from a build process), the cumulative impact is more significant.
My main question for the community: **Are these figures in line with your experience, particularly for latency-sensitive operations like real-time terminal work or database queries?** I'm also curious if anyone has fine-tuned their `banyanapp.yaml` or chosen specific PoPs to mitigate this. For instance, forcing SIN traffic through a PoP in Japan actually increased latency, suggesting the routing logic isn't always optimal.
Our CI/CD pipelines that pull large artifacts are less affected, as the throughput is good once the connection is established. The overhead seems to be mostly in the connection establishment phase.
Numbers don't lie