Werner Vogels on building scalable and robust distributed systems
An explainer on USSD: how the session-based 2G protocol works, why it remains vital in parts of Africa for payments and everyday services on feature phones, and why builders should prioritize simple, appropriate system designs supported by modern cloud/ML/IoT backends.
A senior Amazon developer reflects on how experience improves engineering: recognizing repeating patterns, mentoring juniors, staying skeptical of hype, and using creativity. He discusses the rapid rise of generative AI, advises measured experimentation (pause, learn, ask customers), and notes AWS’s role in democratizing models with safety and privacy guardrails.
An engineering deep-dive on building Amazon Aurora DSQL: the team redesigned the database into modular components (journals, Adjudicator, Crossbar) to scale writes and reads, discovered JVM garbage-collection tail-latency and throughput limits via simulation, and moved the data plane (and eventually the control plane) to Rust for predictable performance, memory safety, and lower tail latency while integrating with PostgreSQL via carefully designed extensions.
A short, reflective reading list urging readers to "think like a fox" — embrace broad, multidisciplinary perspectives to better anticipate technological and social futures. Recommends fiction and nonfiction (Forster, Vonnegut, Doctorow, Malka Older, Tetlock, Lavrijsen, Berlin) that explore automation, network dependence, governance, prediction, and the value of deliberate rest.
A forward-looking, high-level set of technology predictions for 2025+, arguing that the workforce will become increasingly mission-driven; energy innovation (renewables, SMRs) and more efficient, load-aware hyperscale data centers will be critical as compute demand grows; AI and OSINT tools will help combat disinformation; community-sourced open data and edge computing will reshape disaster preparedness; and consumer devices will shift toward intentional, mindful designs.
The article argues that digitalization—especially machine learning, big-data analytics, IoT and voice interfaces—is already reshaping work by automating data-heavy tasks and enabling new human–machine collaboration. Using examples (radiology tools, Amazon Go, Alexa at NASA, rice-farming recommendations), it claims AI will augment and 'rehumanize' work, create new roles and broaden participation, while urging organizations to prepare through skills development and thoughtful change management.
A retrospective on ten years of AWS Availability Zones explaining how physical and logical compartmentalization (zones, regions, cells, and PoPs) have driven service design and operations to reduce blast radius, improve availability, and simplify recovery. The post covers zone-by-zone deployments, regional isolation, asynchronous cross-region features, and deeper intra-zone compartmentalization as continued strategies for reliability and security.
The article argues that data is now at the core of business value and companies must "industrialize" the production of software and data by applying lean manufacturing principles, DevOps and continuous feedback. It highlights the need for cross-functional autonomous teams, automated CI/CD and cloud infrastructure (with examples like SimScale, Kärcher, Breuninger, Autoscout24 and P3) and concludes that organizational change is the main bottleneck to becoming a digital champion.
The article argues that companies can now broadly benefit from machine learning because of abundant digital data, affordable cloud compute, and recent algorithmic advances. It surveys practical business uses — personalization and recommendations, demand forecasting, fulfillment and robotics, document classification, and applications in education and medicine — and highlights that pre-configured libraries, deep learning frameworks and web services make ML accessible to organizations of all sizes. The author also stresses ethical and societal considerations and envisions ML freeing humans from tedious tasks while enabling new business models.
Argues that companies should stop waiting for perfection and instead experiment aggressively, treat failures as learning opportunities, and design anti-fragile systems. Recommends practices such as MVPs, microservices, continuous failure-injection (e.g., Chaos Monkey), root-cause analysis (5 Whys/cause-of-error), and leadership that rewards experimentation to drive faster innovation and more resilient systems.
Summary of a SIGMOD'17 paper describing Amazon Aurora’s architecture: a cloud-native relational database that pushes redo processing to a multi-tenant scale-out storage service to reduce network traffic, enable fast crash recovery and lossless failovers, provide fault-tolerant self-healing storage, and use an efficient asynchronous consensus scheme; includes lessons from 18+ months of production operation.
The article argues that Industry 4.0 — the combination of IoT, cloud connectivity and analytics — is redefining value creation for Germany's manufacturing 'hidden champions.' By connecting shop floors to cloud analytics and combining engineering expertise with digital talent, companies can enable rapid customization, new software-and-service business models, and ecosystems of complementary services. The piece uses examples (Stölzle/Actyx, WATTx) and warns that firms that fail to integrate IT strategically and adopt agile approaches risk losing competitiveness.
A short All Things Distributed post describing the author's weekend reading on fMRI: a brief explanation of how fMRI measures brain activity via blood-flow changes, noting research advances and ethical questions, and listing two review papers (Bandettini 2012; Poldrack & Farah 2015).
A short reading recommendation for Margo Seltzer et al.'s 1993 paper implementing a log-structured file system for UNIX; the post highlights the value of implementing research in production systems, notes practical challenges (citing Dynamo), and points out that this paper measured the implementation's behavior in a production setting.
A short curated recommendation to read Jürgen Schmidhuber’s 2014/2015 overview paper “Deep Learning in Neural Networks” to gain historical context and understand the state of the art in deep learning and neural network algorithms.
Practical guidance on becoming an employer attractive to digital talent: remove rigid silos, form small agile cross-functional teams, empower employees with decision-making and P&L responsibility, provide individualized development paths, create connecting roles (e.g., CDO), hire people with mixed skills and a data-first mindset, and cultivate a culture that tolerates experimentation and meaningful work. Careful, strategic recruiting is essential to successful digital transformation.
A short reading recommendation explaining blockchain basics (a decentralized, immutable, cryptographically time-stamped ledger) and pointing to three 1990s papers — Haber & Stornetta on time-stamping, Schneier & Kelsey on secure logs, and Anderson on decentralized storage — as foundational works that illuminate blockchain's technical roots.
A short reading recommendation of Jim Gray’s 1985 Tandem report “Why Do Computers Stop and What Can Be Done About It?”, which analyzes real-world system failures and introduces software fault-tolerance ideas (process-pairs, transactions) — tracing the shift from hardware to software approaches and noting the paper’s lasting influence on reliability and distributed-system design.
A short blog post recommending three classic papers on the Byzantine Generals problem; it explains the challenge of reaching agreement in the presence of malicious (Byzantine) nodes, notes the oral-message >2/3 loyal threshold and the written-message exception, and links to the foundational SIFT, Reaching Agreement, and Byzantine Generals papers.
A short blog post recommending classic readings on Monte Carlo methods: it explains the idea of using randomness and repeated sampling for numerical problems, gives historical context (Metropolis, Ulam, Manhattan Project), and links to the original 1949 Metropolis & Ulam paper and an Anderson piece on MANIAC.
