Microservices Orchestrator with Locality-Aware Autoscaling
Asty places service instances on nodes closest to real traffic. Requests stay local. Latency stays low. Resources stay efficient.
Traffic in Frankfurt? Services spin up in Frankfurt. No cross-datacenter hops. No wasted latency.
Scales on real validated traffic, not bot noise. Gateway filters junk. Autoscaler reacts to what matters.
mTLS between nodes. API on loopback only. Secrets in env, never in args. Unprivileged process execution.
Agent and Server in one binary. Leader elected automatically via NATS. No external dependencies besides NATS.
Nodes across the globe, one cluster. Geo-diverse redundancy with minimum 3 copies in different DCs.
Canary rollout. Automatic health checks. Auto-revert on failure. Rolling updates one instance at a time.
Gateway runs on every node and reports validated traffic metrics. When real user traffic exceeds the threshold on a node without services, Asty places instances there automatically.
Only authenticated requests that pass rate limiting count toward scaling decisions. Bots and scanners are filtered at the Gateway level and never trigger instance placement.
Baseline of 3 copies per service across different datacenters. If one drops, immediate restoration without cooldown. Geo-diversity always maintained.
Scale up in 30 seconds — users shouldn't wait. Scale down after 5 minutes of calm — no flapping from traffic spikes.
Prometheus metrics, structured logs, health monitoring. Built-in Web UI shows nodes, services, scaling history, and live logs.
All communication over NATS — already in your stack. No extra ports, no extra protocols. State in JetStream KV. Leader election via JetStream.
From bare server to running cluster member in under a minute.