← Back to work
Fault-tolerant distributed task engine · Jun 2026
TaskMesh
A distributed task engine for real-time event processing, coordinated through Kafka and Redis — built to keep working correctly when a worker dies mid-batch, not just when everything goes right.
100K+
req/s across 3 worker nodes
20ms
p99 latency
3
worker nodes, horizontally scaled
How it works
A worker dies mid-batch — at-least-once consumption and idempotent handlers mean the work is redelivered, never dropped and never double-applied.
Scales horizontally by adding consumers to the group; coordination runs through Kafka and Redis.
KafkaRedisDistributed systemsFault tolerance
Fine print
- The harness behind these figures isn't published yet — treat the boundary as unconfirmed.