Open source · Apache 2.0

Watch your infrastructurethink out loud.

DiffOps is an event-driven monitoring platform for distributed systems. Metrics stream in over REST, flow through Kafka, and land as real-time alerts — fault-tolerant by design, cloud-native from the first commit.

events processed today0
How it works

One straight line from metric to alert.

Click a stage to see what it does. Every arrow below is a real network hop — no hidden steps, no magic middleware.

Full architecture
Decouples producers from consumers

Kafka Event Bus

Apache KafkaPartitioned topicsConsumer groups

Validated events land on partitioned topics (metrics.raw, alerts.triggered). Partitioning by service ID preserves per-service ordering while allowing consumer groups to scale horizontally. Retention and replication protect against downstream outages.

Dashboard

Every signal, one screen.

A live look at fleet health — sample data shown, wired to real metrics once the ingestion pipeline is deployed.

diffops.dev/dashboard — production / us-east-1 All systems operational
CPU usage-3.1%
42%
Memory+1.4%
68%
Disk I/O-0.6%
24%
Network+8.2%
312 MB/s
Request rate+2.9%
18.4k/s
Kafka throughput+5.0%
6.1k msg/s

Service status

  • ingestion-api
    14ms99.98%
  • kafka-cluster
    3ms99.99%
  • aggregator-service
    22ms99.95%
  • alert-engine
    61ms99.71%
  • postgres-primary
    5ms99.99%
  • redis-cache
    1ms100%

Recent alerts

  • alert-engine p95 latency above 50ms for 5m
    2m ago
  • aggregator-service scaled 3 → 5 pods
    18m ago
  • us-west-2 consumer lag exceeded 10k messages
    1h ago · resolved
Request lifecycle

What happens between a metric and an alert.

Seven hops, end to end — typically under 400ms from emission to dashboard.

  1. 01

    Emit

    An agent on the host batches CPU, memory, disk, and network metrics every few seconds.

  2. 02

    Ingest

    The REST gateway authenticates the batch, validates its schema, and hands it to Kafka.

  3. 03

    Stream

    Kafka partitions the event by service ID and holds it for every interested consumer.

  4. 04

    Process

    Consumer services enrich, aggregate, and compute rolling percentiles in near real time.

  5. 05

    Persist

    Aggregates land in PostgreSQL for history; the hot window lives in Redis for instant reads.

  6. 06

    Evaluate

    The alert engine checks the fresh window against your rules on every incoming event.

  7. 07

    Notify

    Breached thresholds fire an alert, deduplicated and rate-limited before it reaches you.

Use cases

Where DiffOps fits.

Microservice fleets

Watch dozens of independently deployed services without wiring a new dashboard for each one.

Self-hosted infrastructure

Run DiffOps alongside your own bare-metal or VM fleet where managed monitoring isn't an option.

Event-driven systems

If your architecture already speaks Kafka, DiffOps slots in as another well-behaved consumer.

Capacity planning

Historical throughput and saturation trends make it obvious when to scale — before it's an incident.

Roadmap

Where DiffOps is headed.

Full roadmap
  • Core ingestion pipeline
    Q2 2026
  • Live operational dashboard
    Q3 2026
  • Auth, workspace isolation, and RBAC
    Q4 2026
  • Alert routing and notification channels
    Q1 2027
  • Kubernetes operator and Helm support
    Q2 2027
  • Prometheus / Grafana integration
    Q3 2027
Early feedback

What early adopters are saying.

The producer-consumer split made it trivial to scale just the aggregation layer during a traffic spike, without touching ingestion at all.
Platform engineer, early adopter
Reading the Kafka topic layout took ten minutes and I understood the whole data flow. That's rare for a monitoring stack.
Backend developer, early adopter
Self-hosting it next to our existing cluster was the deciding factor — no external agent, no data leaving our VPC.
Infrastructure lead, early adopter
FAQ

Common questions.

Yes. DiffOps is fully open source under the Apache 2.0 license — self-host it, fork it, or modify it for your own infrastructure at no cost.

Bring your infrastructure into view.

Spin up the full stack locally in a few minutes with Docker Compose.