http://ddinsights.net/tags/fundamentals/Recent content in Fundamentals on Distributed Data InsightsHugoen-usSun, 07 Jun 2026 10:00:00 -0700http://ddinsights.net/posts/why-distributed-data-is-hard/Sun, 07 Jun 2026 10:00:00 -0700http://ddinsights.net/posts/why-distributed-data-is-hard/<p>Every distributed data system is an answer to the same uncomfortable question: <em>what do you do when part of your system fails but the rest keeps running?</em> On a single machine, a crash takes everything down together — clean, if catastrophic. Across a network, failure is <strong>partial</strong>, <strong>ambiguous</strong>, and <strong>constant</strong>.</p> <h2 id="the-three-things-the-network-takes-away">The three things the network takes away</h2> <p>When you split state across machines, you lose three guarantees you took for granted on a single box:</p>http://ddinsights.net/posts/lsm-trees-vs-btrees/Tue, 02 Jun 2026 08:30:00 -0700http://ddinsights.net/posts/lsm-trees-vs-btrees/<p>Underneath nearly every database is one of two storage engines: a <strong>B-tree</strong> or an <strong>LSM-tree</strong>. The choice shapes write throughput, read latency, space usage, and how the system behaves under pressure. It&rsquo;s worth understanding <em>why</em> the big distributed stores — Cassandra, RocksDB, HBase, ScyllaDB — overwhelmingly chose LSM-trees.</p> <h2 id="b-trees-update-in-place">B-trees: update in place</h2> <p>A B-tree keeps data sorted in fixed-size pages and <strong>mutates pages in place</strong>. To update a key, you find its page and overwrite it. Reads are excellent — typically a handful of page lookups — and the structure has powered relational databases for decades.</p>