June 2025
- Jun 12, 2025ExplainersHow Database Indexes Improve Query PerformanceLeis and colleagues evaluated PostgreSQL alongside commercial systems and found estimation errors drove the planner to pick query plans 10 to 100 times slower than optimal, so a correct index can still go unused. The structural case for indexes remains overwhelming: with a branching factor of 500, a B-tree reaches 62.5 billion rows in four levels. This page covers clustered versus non-clustered, covering indexes, BRIN, the write penalty and reading EXPLAIN output.
- Jun 12, 2025ExplainersUnderstanding Container Technology and Its BenefitsThe Borg paper by Verma and colleagues describes bin packing across clusters of machines, and resource efficiency, not developer convenience, was the primary motivation. That inverts the usual framing of the Docker era. This page traces the lineage from chroot in 1979 through FreeBSD jails and Solaris Zones, which added resource controls years before cgroups, then covers namespaces, layered images and why a shared kernel is a weaker boundary than a hypervisor.
- Jun 12, 2025ExplainersHow Version Control Systems WorkWhat actually happens when you commit: SHA-1 content addressing that deduplicates identical blobs, the four object types, branches that are nothing more than a 41-byte file holding a commit hash, and the three-way merge that uses a common ancestor to decide what conflicts. The research sections add DORA's finding that elite teams merge branches within a day while low performers keep them alive for weeks.
- Jun 11, 2025ExplainersHow APIs Work: Technical Deep DiveRoy Fielding's 2000 dissertation at UC Irvine defined REST as six constraints, and named statelessness the one that matters most for scaling: because no server holds session data, any instance can serve any request. Yet the Richardson Maturity Model, introduced in 2008, shows most production APIs stop at Level 2, using HTTP methods and status codes properly while skipping hypermedia links. This walks the request path and the choices behind REST, GraphQL, gRPC and SOAP.
- Jun 11, 2025ExplainersUnderstanding Database Transactions and ACID RulesTransferring $500 means debiting one account and crediting another. If the system crashes between the two, the money either vanishes or is created from nothing, and both break the invariant that total money stays constant. A transaction wraps the writes so they all commit or all roll back. ACID, coined by Theo Haerder and Andreas Reuter in 1983 on Jim Gray's work, names four interlocking guarantees: remove one and corruption, violated constraints or lost commits follow.
- Jun 10, 2025ExplainersHow Load Balancers Distribute TrafficMichael Mitzenmacher's power-of-two-random-choices result cut the maximum load on any server from O(log n / log log n) under plain random selection to O(log log n), and now ships inside Nginx and Envoy. The article walks through each algorithm's state cost, Layer 4 versus Layer 7 trade-offs, active and passive health checks, sticky-session penalties, and the three TLS strategies: termination, passthrough and bridging.
- Jun 10, 2025ExplainersHow Encryption Protects DataSecurity comes from the key, not from hiding the method: Kerckhoffs stated in 1883 that a system should stay secure even if everything but the key is public, so AES, RSA and TLS are all published openly. TLS uses asymmetric cryptography only for the handshake, then switches to symmetric encryption for speed. Enigma offered roughly 158 quintillion settings and still fell to procedural weaknesses, which is why key management, not algorithm strength, is the usual failure point.
- Jun 10, 2025ExplainersThe Process of DNS Resolution UncoveredThe fix widened the space an attacker must guess and made brute-force cache poisoning roughly 65,000 times harder. DNS replaced a single hand-edited HOSTS.TXT file at SRI that was updated once or twice a week, after Mockapetris published RFC 882 and 883 in November 1983. This page traces a query from browser cache through the root and TLD servers to the authoritative answer, then covers TTLs, record types, DNSSEC, DoH and the 2016 Dyn attack.
- Jun 9, 2025ExplainersAn Overview of Machine Learning Model TrainingKrizhevsky, Sutskever and Hinton won using ReLU activations, dropout and GPU acceleration, techniques that were already available but had not been combined this way. That is the real lesson of 2012. This page works through the training loop from first principles: loss functions, gradient descent and the Adam optimizer, backpropagation and vanishing gradients, overfitting and regularization, why data is split three ways, and how to tell when training is finished.
- Jun 9, 2025ExplainersHow Search Engines Crawl, Index, and Rank ContentA single search draws on a distributed inverted index spanning millions of servers, then retrieves, scores and ranks candidates across hundreds of signals in roughly 200 milliseconds. The web began as hand-curated directories that could not scale past tens of thousands of pages, then keyword engines trivially gamed by hidden text. Page and Brin's 1998 paper introduced PageRank, treating a link as a vote weighted by the voter's own importance.
No guides match that search. Try a shorter phrase, or clear the search.