Show-off or self-reflection? Somewhere in between. I care more about the process of understanding a system than the result of shipping it. If that sounds counter-productive, you might want to skip me over.
Consensus · replication · failure detection · partitioning & sharding · event-driven & CQRS messaging · log-structured storage.
- EastGuard – Zero-controller distributed messaging built for the scale that breaks Kafka
- Metadata sharded across many small Raft groups, so coordination throughput scales with the cluster instead of funneling through a single controller
- SWIM gossip for decentralized, O(log N) failure detection — no central heartbeats, no heartbeat storm
- Dynamic ranges that split and merge with load; segments striped across nodes for automatic balancing
- Seal-on-failure replication in place of Kafka's ISR — seal the segment, reopen on a healthy replica set, treat sealed data as immutable
- Duva – Strongly-consistent distributed key-value store
- Replicated storage designed around correctness under node and network failure
- Ruva – Event-driven framework
- Message-driven core, async end to end
- CQRS / Event-Sourcing Framework
- Command/query separation with an append-only event log as the source of truth
- IAM — identity & access management service
- Event-driven architecture
- Domain-Driven Design
- Async-first
- GitOps Setup
- Kubernetes Basics
- Docker, ArgoCD, SealedSecrets, LetsEncrypt
