field notes // databases, storage engines & the systems beneath them

Notes from inside the machine, line by line.

Long-form dives into how databases, storage engines, and the systems underneath them actually work, from the internals, not just the docs.

Start Reading browse the log

wal.log (tail -f)

0141APPENDkey=user:8841 seg=012

0142APPENDkey=session:229f seg=012

0143COMMITtxn=77a3

0144FLUSHmemtable → sstable-0091.sst

0145CHECKPTwal_truncate=true

Latest Entry

LSN 011

NetworkingLSN 011

What Happens When a Backend Dies Mid-Request?

Health checks and retries sound like the same feature. Building both into a Go reverse proxy and timing a real backend failure showed me they solve two completely different problems.

Rasika Lakmal·2026-10-12·8m

Read Article

The Log

11 entries

NetworkingLSN 010

Round Robin vs Least-Connections, Benchmarked Instead of Assumed

Two load-balancing strategies that sound interchangeable until the backends behind them aren't the same speed, tested with real numbers from a Go reverse proxy I built myself.

2026-10-11 · 7m

NetworkingLSN 009

Why Connection Pooling Matters More Than It Sounds Like It Should

I built a reverse proxy in Go, then built a second, deliberately naive one with no connection pooling, and benchmarked them side by side to see what pooling actually buys you.

2026-10-10 · 8m

NetworkingLSN 008

TCP Connections vs HTTP Requests: Why a Byte-Copying Proxy Works Until You Need Per-Request Anything

Building a reverse proxy from scratch in Go, starting with a version that just copies bytes and understanding exactly why that stops being enough.

2026-10-09 · 7m

DatabaseLSN 007

B-Tree vs LSM-Tree: What My Own Benchmarks Actually Showed

The classic storage-engine tradeoff, tested with real numbers from a database I built myself, and an honest note about what wasn't actually measured.

2026-10-02 · 9m

DatabaseLSN 006

Why Databases Need a Write-Ahead Log

The deceptively simple question 'why not just write directly to your data files?' answered by building a write-ahead log from scratch in Rust.

2026-10-01 · 7m

DatabaseLSN 005

Memtables and SSTables, Explained by Building Them

Why RocksDB, Cassandra, LevelDB, and ScyllaDB all reach for the same two data structures, explained by building a small LSM-tree database from scratch in Rust.

2026-09-30 · 8m

DatabaseLSN 004

What Happens When Your Database Crashes Halfway Through a Write?

Building a small LSM-tree database from scratch in Rust and deliberately killing it, corrupting it, and torn-writing it to find out what 'crash-safe' actually means.

2026-09-29 · 9m

NetworkingLSN 003

What a Speed Test Is Actually Measuring

Building a small Go backend for measuring network speed taught me that the hard part isn't measuring bytes, it's getting buffering, compression, caching, and timeouts out of the way first.

2026-02-27 · 7m

ObservabilityLSN 002

Integrating OpenTelemetry with NestJS and Grafana Tempo

Step-by-step guide to integrate OpenTelemetry (OTel) with a NestJS app, run Grafana and Tempo with Docker, and visualize traces.

2025-09-20 · 5m

DatabaseLSN 001

How to Analyze and Optimize PostgreSQL Query Cost Using EXPLAIN (pgAdmin + Dalibo)

Master PostgreSQL performance tuning by learning how to analyze and optimize query costs using pgAdmin's EXPLAIN and visualizing execution plans with Dalibo Explain.

2024-02-11 · 6m