About Systems Explained

Distributed systems are hard to understand from documentation alone. You can read about Raft consensus, LSM-trees, or partition rebalancing — but until you see them in action, the tradeoffs don't click.

The Problem

Most distributed systems education is either too theoretical (academic papers) or too practical (vendor docs that assume you already understand the concepts). There's a gap in the middle: interactive, visual explanations that let you play with the parameters and see what happens.

Our Approach

Every visualization on this site is designed around a simple principle: feel the tradeoffs, don't just read about them.

• Drag a slider to change region size — watch parallelism, overhead, and recovery time respond
• Click to add data — see it flow through LSM-tree levels and trigger compaction
• Simulate a node failure — observe how the system recovers (or doesn't)

Who We Are

We're engineers who've spent years operating distributed databases, message queues, and compute clusters in production. We built these visualizations because they're the explanations we wished existed when we were learning.

How pages are researched

Every system-level claim is grounded in primary sources: the project's source code, official documentation, kernel docs, RFCs, or the original academic paper. When we say PostgreSQL uses MVCC with a transaction ID that wraps around at 2^32, that number comes from src/backend/access/transam/varsup.c, not from a third-party tutorial. When we say a Raft cluster of N nodes survives ⌊(N-1)/2⌋ failures, that's from the original Raft paper (Ongaro & Ousterhout, 2014), section 5.4.

Where systems behave differently in practice than in their specification — for instance, Kafka's acks=all not actually waiting for all replicas if the in-sync replica set has shrunk — we explain both the spec and the operational reality, with the relevant config knob that controls the difference.

Editorial principles

• Verified, not summarized. Every mechanic on the site is traced back to the source — code, docs, or paper — not to another blog post.
• Operational, not academic. The level of detail aims at engineers making production decisions, not students passing exams.
• No sponsored content. Pages are not edited in exchange for payment, and protocol comparisons are not influenced by vendor relationships.
• Errors get fixed publicly. If a page is wrong, please file a correction. Substantive errors get an inline correction note explaining what changed.

Open Source

The visualizations on this site are open source. If you spot an error, want to add a feature, or have an idea for a new visualization, contributions are welcome on GitHub.

Topics covered

The site groups content into a dozen broad areas: PostgreSQL, ClickHouse, DuckDB, and other database internals; the Linux kernel (eBPF, /proc, scheduler, page cache, the I/O stack); Kubernetes and container runtimes; networking (TCP, HTTP/2 and HTTP/3, DNS, TLS); distributed-systems primitives (Raft, Paxos, gossip protocols, partition strategy); Kafka and streaming; LLM internals (transformers, KV cache, inference stacks); system-design interview problems (URL shortener, rate limiter, distributed cache, social-graph design); and security (OAuth, JWT, TLS handshake, zero-trust). Each area has a hub page that links the canonical deep-dives in that space — see the homepage for the full topic list.