How Symbiosis routes your traffic
Observe, predict, shift: three steps that run automatically from the moment you deploy.
Three steps, fully automated
No human in the loop unless you want one.
Observe
Every PoP streams telemetry back to the control plane every 500ms: requests per second, error rate, and latency percentiles. The control plane builds a rolling baseline for your application's traffic shape.
Predict
An ML model trained on your own traffic history flags emerging spikes before they peak. No generic thresholds. No manual tuning. The model learns your patterns from day one.
Shift
Capacity moves to the right PoP before the wave arrives. Containers pre-warm, routing weights adjust, and your users reach a warm instance. Average lead time across early-access pilots: 40 seconds.
40-second lead time
The prediction model uses your app's own telemetry history. No configuration required. It detects pattern signatures that typically precede a spike and triggers a capacity shift before the first request hits a cold instance.
- 40s average lead time across early-access pilots
- Zero missed spike events in the first cohort's 90-day window
- Model retrained on your own history, not generic baselines
12 PoPs across 5 continents
Every region routes requests to the nearest live node. Latency stays low regardless of where your users are.
| City | Region | Status |
|---|---|---|
| Stockholm | Europe North | Live |
| London | Europe West | Live |
| Amsterdam | Europe West | Live |
| Frankfurt | Europe Central | Live |
| Paris | Europe West | Beta |
| New York | North America East | Live |
| Chicago | North America Central | Live |
| Los Angeles | North America West | Live |
| Sao Paulo | South America | Beta |
| Singapore | Asia Pacific | Live |
| Tokyo | Asia Pacific | Live |
| Sydney | Asia Pacific | Beta |
Every deploy gets its own dashboard
No extra configuration. The moment you push, Symbiosis starts collecting request counts, error rates, and per-region latency breakdowns. You see exactly what is happening at each PoP in real time.
- Request bar chart, 24-hour rolling window
- Error rate chip with 5-minute and 1-hour buckets
- Top 5 PoPs by volume with region breakdown
Four commands. Global reach.
No YAML pipeline required. The Symbiosis CLI connects to your repository, detects pushes via webhook, and handles distribution to all PoPs automatically.
- Works with any Git host: GitHub, GitLab, Bitbucket
- Zero changes to your build process
- Rollback to any prior deploy with one command
Platform FAQ
-
The model starts building your baseline from the first hour of traffic. It reaches useful accuracy after roughly 3-5 days of consistent activity. Until then, Symbiosis falls back to threshold-based alerts so no spike goes unhandled.
-
Request count, error rate, and p50/p95/p99 latency per PoP, sampled every 500ms. No request body content is collected. You can export all telemetry to Datadog or Prometheus via the integration layer.
-
Yes. Each linked repository is treated as a separate project with its own PoP routing configuration, dashboard, and prediction model. Projects share the account's request quota.
-
The control plane detects an unhealthy PoP within one health-check cycle (typically under 10 seconds) and reroutes traffic to the next-nearest live node. You receive an alert via whichever notification channel you have configured.
Start with Hobby. Zero commitment.
Free tier gets you one PoP and 100K requests per month. Scale when you need it.