PoCAT Documentation Get Started

Gateway Architecture Overview

PoCAT Gateway five-stage pipeline, instance sizing, and operational KPIs.

Last updated: 2026-05-27 Section: Architecture

PoCAT Gateway is the single entry point for enterprise integration. Every request passes through Ingress → Route Resolve → Filter Chain → Servit Execution → Egress; each layer must be observable, scalable, and fault-isolated independently.

Before you begin Understand deployment topology (Kubernetes/Swarm) and XML paths (config/, routes/), then adopt the KPI/SLO tables in this guide as team standards.

Pipeline layers

LayerCore responsibilityOps KPIFailure signalRecommended action
IngressProtocol intake, parsing, basic validationRPS, 4xx ratioAccept backlog growthScale thread/connection pools
Route ResolveRule matching, upstream selectionMatch success rateFallback route spikeReview priority and rule diffs
Filter ChainAuth, validation, transform, auditFilter latency p95401/403 or filter delayCheck chain order and cache
ServitDomain logic executionSuccess ratio, P95Business error code spikeValidate upstream and schema
EgressExternal system callsTimeout, retry countDownstream timeout spikeTune CB, queue, timeouts

Size options (Gateway instances)

Scale JVM heap, threads, and connection pools by traffic tier. Production should use Medium or larger.

SizeApprox. RPSJVM HeapReplica (prod)Notes
Small~5002 GiB2Dev/staging
Medium~2,0004 GiB3Typical production
Large~8,0008 GiB3+HPA enabled
Extra Large8,000+16 GiB+5+Dedicated nodes, anti-affinity

Update gateway sizing

  1. Collect P95 latency, CPU, and heap over a 7-day window.
  2. Pick size with ~20% burst headroom over peak RPS.
  3. Update resource and pool settings in config/gateway-config.xml.
  4. Load-test in staging, then roll out with maxUnavailable: 0.
  5. After deploy, verify error rate and P95 stay within SLO for 30 minutes.

Recommended design principles

  • Keep security policy in filters; handle domain policy in Servits.
  • Manage routing declaratively in XML—avoid hard-coding in application code.
  • Generate a Correlation ID per request and propagate it downstream.