Built for scale & speed

A modular, event-driven architecture that processes millions of signals per second while keeping query latency under 100ms.

LAYERSFour-Layer Architecture

01

Ingestion Layer

OTLP gRPC & HTTP + native JSON API. Accept traces, metrics, and logs from any language via OpenTelemetry.

Port 4317 (gRPC) · Port 4318 (HTTP)
02

Processing Pipeline

Kafka-backed pipeline for aggregation, anomaly detection, topology extraction, alert evaluation, SLO compliance, reliability scoring, and synthetic monitoring.

Go microservices · Kafka · Redis
03

Storage Layer

Purpose-built stores for each signal type. ClickHouse for traces & logs, VictoriaMetrics for metrics, Neo4j for topology.

ClickHouse · VictoriaMetrics · Neo4j · PostgreSQL
04

Intelligence Engine

Python AI engine for root cause analysis, anomaly detection, baseline generation, predictive analytics, and capacity forecasting.

PyTorch · scikit-learn · Prophet · LSTM

SERVICESCore Microservices

Ingestion Service

OTLP gRPC/HTTP receiver with validation and Kafka publishing

Processor

Kafka consumer writing to ClickHouse, VictoriaMetrics, and running scheduled jobs

API Gateway

REST + GraphQL + WebSocket on :9080 with JWT auth, RBAC, and rate limiting

Topology Service

Trace-to-graph extraction with Neo4j for service dependency mapping

Alerting Engine

PromQL and trace-based rules evaluated every 15s with M365 and Teams notifications

Healing Engine

Policy-based auto-remediation with approval workflows and rollback support

AI Service

gRPC on :50051 — anomaly detection, RCA, baselines, prediction, and forecasting

DATAStorage Infrastructure

Kafka

Event streaming connecting all pipeline stages

ClickHouse

Columnar storage for traces and logs

VictoriaMetrics

Time-series database for metrics

PostgreSQL

Users, roles, config, policies, audit trail

Redis

Cache, rate limiting, session store

Neo4j

Graph database for service topology

Ready to see everything?

The Pointer team handles deployment and onboarding end-to-end. Request a demo and we'll have you monitoring in days, not weeks.