Serge Skoredin — IT Blog

What actually holds up when you build observability into a Go service: what belongs in a metric vs a log vs a trace, a measured cost of getting that wrong, dashboards and alerting that don't cause fatigue, and PII discipline - not another setup tutorial.

Updated with real go test -bench numbers from two machines (Apple M5, Hetzner AMD EPYC), after Hacker News pushed back on several unverified claims.

encoding/json, jsoniter, easyjson, sonic, and Go 1.27's encoding/json/v2 benchmarked head to head on the same payload, on an Apple M5 and a Hetzner VPS - including a jsoniter decode result that only holds up on one of the two machines.

Comparison of PostgreSQL, MongoDB, Cassandra, and ClickHouse.

Everyone talks about clean architecture.

Memory usage: 47GB.

Master all four gRPC streaming patterns.

Build production-ready GraphQL servers in Go.

Build production gRPC Gateway handling 100K+ requests/day.

Microservices aren't your only option.

Production bug from lost orders.

Production feature flags system in Go handling 50M+ evaluations per day.

Event sourcing: append-only log as database.

Production eBPF programs in Go.

Everything Go developers actually need to know about Docker.

Ring cache (circular buffer) explained for HFT: lock-free implementation in Go that achieves sub-microsecond latency.

Real-world guide to rolling out distributed tracing in Go with OpenTelemetry and Grafana Tempo: why logs were not enough, how to instrument services, and what changed once the traces shipped.

Real-world experience with ClickHouse ReplacingMergeTree pain points and modern stream-based deduplication solutions that actually work in production.

Production distributed lock manager in Go.

Building real-time data pipelines with Debezium CDC and Go.

Skip the 500-page books.

Complete guide to E2E encryption implementation.

Building a production-grade WAF in Go.

Built a DSL in Go that handles 10M rule evaluations per day.

Complete guide to building microservices in Go. gRPC, Docker, Kubernetes, and production best practices.

Build production CDN in Go handling 100GB+ daily traffic.

Everyone's building a blockchain.

Built production API Gateway in Go handling 1M+ requests/day.

Why your Go code is slow: AI writes over-engineered solutions, humans copy-paste without understanding.

Tested Claude, Copilot, Cursor, and Codeium on real Go tasks, plus how we actually run AI as a multi-agent pipeline instead of one prompt.

Practical experience integrating GPT-4 into backend services.