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.
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.
Kafka Event Bus
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.
Every signal, one screen.
A live look at fleet health — sample data shown, wired to real metrics once the ingestion pipeline is deployed.
Service status
- ingestion-api14ms99.98%
- kafka-cluster3ms99.99%
- aggregator-service22ms99.95%
- alert-engine61ms99.71%
- postgres-primary5ms99.99%
- redis-cache1ms100%
Recent alerts
- alert-engine p95 latency above 50ms for 5m2m ago
- aggregator-service scaled 3 → 5 pods18m ago
- us-west-2 consumer lag exceeded 10k messages1h ago · resolved
What happens between a metric and an alert.
Seven hops, end to end — typically under 400ms from emission to dashboard.
- 01
Emit
An agent on the host batches CPU, memory, disk, and network metrics every few seconds.
- 02
Ingest
The REST gateway authenticates the batch, validates its schema, and hands it to Kafka.
- 03
Stream
Kafka partitions the event by service ID and holds it for every interested consumer.
- 04
Process
Consumer services enrich, aggregate, and compute rolling percentiles in near real time.
- 05
Persist
Aggregates land in PostgreSQL for history; the hot window lives in Redis for instant reads.
- 06
Evaluate
The alert engine checks the fresh window against your rules on every incoming event.
- 07
Notify
Breached thresholds fire an alert, deduplicated and rate-limited before it reaches you.
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.
Where DiffOps is headed.
- Core ingestion pipelineQ2 2026
- Live operational dashboardQ3 2026
- Auth, workspace isolation, and RBACQ4 2026
- Alert routing and notification channelsQ1 2027
- Kubernetes operator and Helm supportQ2 2027
- Prometheus / Grafana integrationQ3 2027
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.”
“Reading the Kafka topic layout took ten minutes and I understood the whole data flow. That's rare for a monitoring stack.”
“Self-hosting it next to our existing cluster was the deciding factor — no external agent, no data leaving our VPC.”
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.