The article details Xiaomi Push Service's XMPP‑based protocol, multi‑layer server architecture, client SDK design, scalability measures, security mechanisms, performance metrics during the 11.11 promotion, major refactorings, and practical insights for building high‑throughput push systems.
This article shares the architecture, design principles, challenges, and performance‑optimizing solutions behind a ten‑hundred‑thousand‑scale server monitoring system, covering data collection agents, distributed pipelines, real‑time alerts, high throughput, multi‑platform support, and practical lessons learned.
This article details Haier's e‑commerce platform architecture, explaining how a SOA approach built on Dubbo enables high scalability, a flexible product service layer with a DSL query language, and performance optimizations that handle massive traffic spikes during major shopping events.
This article explains how JD.com’s Order Fulfillment Center (OFC) was built, re‑engineered, and continuously optimized to handle massive order volumes during major sales events, covering its architecture, migration from .Net to Java, distributed task queues, flow control, and operational practices that ensure reliability and scalability.
This article examines the technical challenges of massive web sites—such as billions of users, extreme concurrency, and petabyte‑scale data—and explains how architectural evolution, including service separation, caching, clustering, load balancing, CDN, distributed storage, NoSQL, and micro‑service decomposition, enables scalable, highly available, and secure operations.
This article examines the inherent drawbacks of Redis Cluster—such as gossip overhead, upgrade challenges, client protocol complexities, and implementation gaps—and proposes a set of architectural enhancements, including proxy layers, dashboards, and agents, to improve scalability, reliability, and operational efficiency.
During a 2015 tech salon, Weibo platform architect Wei Xiangjun dissected the essence of system architecture, illustrating how Sina Weibo scales to billions of users through multi‑layered services, sharding, CDN, multi‑datacenter caching, orthogonal decomposition, and robust monitoring, offering practical insights for large‑scale backend design.
The article explains the internal and external drivers behind website architecture evolution, outlines the U‑N‑S‑C model, describes scaling steps such as service separation, caching, clustering, load balancing, database read/write splitting, CDN, and highlights the operational and organizational challenges of each phase.
This article summarizes Chen Ming’s 2015 ArchSummit talk on WeChat Moments, detailing its massive posting volume, core database tables, publishing and reading workflows, and the multi‑data‑center disaster‑recovery architecture that ensures seamless, loss‑less service worldwide.
This article examines the push and pull models used by micro‑blogging platforms such as Twitter and Sina Weibo, analyzes their scalability challenges, and proposes a time‑partitioned pull approach that reduces database load while maintaining fast feed retrieval for active users.
This article presents a pattern‑based approach to performance optimization, describing common degradation anti‑patterns and corresponding optimization patterns—such as horizontal and vertical partitioning, runtime 3NF, data locality, and degradation—to help engineers improve response time, throughput, and availability in large‑scale backend services.
From a single‑machine “all‑in‑one” setup to a distributed, service‑oriented architecture, 58.com’s journey illustrates how startups can evolve their backend systems—adopting LAMP, read/write splitting, vertical splitting, caching, Java migration, and automated services—to handle traffic growth from tens of thousands to over a billion visits.
This article chronicles the architectural journey of JD’s Dongdong instant‑messaging platform, detailing its early simple polling design, subsequent performance and scalability challenges, and the progressive shifts toward service‑oriented, micro‑service, and cloud‑native architectures that support massive user growth.
The article presents a pattern‑based catalog of performance‑optimization techniques—such as horizontal and vertical partitioning, constant‑variable separation, locality, and sandboxing—illustrated with Meituan case studies, explaining motivations, trade‑offs, and how each pattern reduces latency, boosts throughput, and enhances availability in high‑throughput backend systems.
Facing massive traffic spikes during events like Double‑11, Haier’s e‑commerce platform adopted a SOA architecture built on Dubbo, designed a flexible product service with a DSL‑driven query engine, and achieved high performance, linear scalability, and extensibility across diverse product categories.
The article chronicles a developer’s journey from a simple personal homepage to managing billion‑visit web platforms, sharing practical insights on architecture evolution, knowledge structures, design principles, infrastructure, and software engineering practices essential for building and operating large‑scale websites.
This article recounts Zhihu's evolution from a tiny Python‑Tornado service on a single Linode to a massive, highly available backend employing custom logging, event‑driven processing, page‑render optimizations, and a service‑oriented architecture that now supports over 100 million users.
This article explores the unique characteristics of e‑commerce search engines, their specialized architecture, core modules, data update processes, and practical solutions for bugs, high concurrency, caching, and cold‑start challenges, offering a comprehensive guide for building robust search systems.
This article examines 58.com’s BangBang system architecture, detailing its four‑layer backend design, evolution from a traditional IM to a mobile management platform, and the push technologies used to support massive O2O traffic and millions of concurrent users.
During Double 11, JD.com’s product detail page handled tens of billions of requests without downtime by introducing a unified service architecture that consolidates promotion, inventory, recommendation, and other functions behind a single Nginx+Lua entry point, employing multi‑level caching, service isolation, and flexible degradation mechanisms.
In this interview, Xiaomi R&D architect Ouyang Chen shares the design philosophy, evolution stages, core modules, pitfalls, and future challenges of the Xiaomi advertising platform, offering practical insights for building scalable backend systems.
This article examines Uber's rapid 38‑fold growth and the engineering choices behind its real‑time market platform, detailing the scheduling system, geographic indexing, microservices, Ringpop, TChannel, and strategies for scalability, availability, and fault tolerance.
This article traces the evolution of large‑scale website architectures—from single‑server setups to distributed services—detailing how separation of concerns, caching, load balancing, database sharding, CDN, reverse proxies, NoSQL, and service decomposition collectively improve performance, scalability, and reliability.
This article details Alibaba's mobile infrastructure evolution that powered a record‑breaking 2015 Double‑11 sale, covering client container design, network pipeline challenges, dynamic rendering, cross‑platform solutions like Weex, and the open‑source plans that enable massive scalability.
This article explains why microservices outperform monolithic architectures by detailing their design, deployment, benefits, and challenges, helping developers decide if the approach suits complex, evolving applications; it also compares traditional single‑code‑base models, discusses scaling strategies, and outlines practical considerations for implementation.
JD’s Order Fulfillment Center (OFC) evolved from a small data‑transfer team into a highly scalable, distributed architecture that handles massive order volumes during events like 618 and Double‑11, employing Java migration, service decomposition, flow control, and robust operations to ensure data consistency and rapid delivery.
This article explains how to design a high‑availability flash‑sale system by separating business and data layers, using Redis queues for pressure isolation, ensuring inventory consistency, handling transaction reconciliation, and addressing anti‑fraud measures to cope with massive concurrent traffic.
The article details how LinkedIn has massively scaled Kafka usage over several years, addressing quotas, a new ZooKeeper‑free consumer, reliability enhancements, security features, monitoring frameworks, fault testing, and ecosystem integrations to support its massive data‑driven operations.
This article outlines the typical progression of website architecture from a single‑server setup through service separation, caching, clustering, read/write splitting, CDN, reverse proxy, distributed storage, NoSQL, and business‑level service decomposition, providing a clear roadmap for scaling web systems.
The article describes how the limitations of an in‑memory NoSQL HOLD platform led to the creation of TDSQL, a distributed MySQL‑based SQL system featuring ZooKeeper‑coordinated Scheduler, Agent, and Gateway modules, automatic cross‑IDC disaster recovery, transparent horizontal scaling, strong synchronous replication, and future integration with container technologies.
The article chronicles how 58.com’s web architecture evolved from a tiny, single‑machine setup to a multi‑layer, Java‑based, highly available service‑oriented platform, detailing the technical decisions, scaling challenges, and automation practices adopted at each traffic milestone.
This article explains what system load capacity is, how to measure it using requests per second, identifies key factors such as bandwidth, hardware, OS and server configurations, and provides practical tuning steps for Linux, TCP, Nginx, Tomcat and databases to improve scalability.
This article details how 58.com’s Bangbang system evolved from a traditional IM platform to a massive, multi‑layered architecture handling over 40 billion daily requests, covering its four‑layer design, third‑party integration challenges, client lightweighting, and the push technologies that power its mobile O2O services.
This article examines the core challenges of massive web platforms—huge user bases, high concurrency, and massive data—and walks through a step‑by‑step evolution of architecture from a single‑server setup to distributed caching, clustering, read/write separation, CDN, and micro‑service strategies for achieving high availability, performance, and security.
This article chronicles the step‑by‑step evolution of Weibo's recommendation system architecture—from the early independent 1.0 stage through the layered 2.0 phase to the current platform‑centric 3.0 design—detailing the driving business and technical factors, architectural components, advantages, shortcomings, and key outcomes at each stage.
This article outlines fifteen common hidden requirement types—from detail tweaks and cross‑platform adaptations to scalability, security, and content‑operation needs—and offers practical strategies for developers and product teams to anticipate, design for flexibility, and reduce costly changes during development.
The article analyzes the inherent drawbacks of Redis Cluster—such as its decentralized P2P design, gossip overhead, upgrade difficulty, lack of hot‑cold data separation, client protocol challenges, and implementation limits—and proposes architectural enhancements like proxy, dashboard, and agent components to improve scalability, manageability, and performance.
The article chronicles Tencent's technological evolution across four stages—from massive user growth to flexible platforms—highlighting the strategic use of efficient architectures, dynamic operations, and cloud services that enabled it to support billions of users while maintaining performance and reliability.
This article details JD.com's transition from static to dynamic, high‑performance product detail page architecture, describing the multi‑layer system design, data heterogeneity, caching strategies, scaling techniques, operational challenges, and the solutions implemented to meet rapid business demands during massive traffic events.
An in‑depth look at 360’s long‑connection push platform built with Go, covering system architecture, key performance indicators, scalability challenges, optimization strategies, operational practices, and practical Q&A for building reliable, high‑throughput messaging services.
The article explains how distributed systems split a task into multiple subtasks across separate servers to reduce execution time, while clusters replicate the same service on many servers to increase throughput, highlighting their complementary roles in scalability and high availability.
Vertical sharding splits large tables across multiple servers by whole tables, offering simple partitioning but limited scalability, while horizontal sharding distributes rows of high‑traffic tables based on key rules, improving performance and scalability yet introducing complex partition logic and maintenance challenges.
The article explains the historical background, core characteristics such as scalability, cost‑efficiency, fault tolerance, and elasticity, and outlines essential design principles for distributed systems, emphasizing low hardware requirements, horizontal scaling, avoidance of single points of failure, reduced communication overhead, and stateless services.
This article compares relational databases with distributed document-oriented NoSQL databases, explaining their scaling models, data modeling differences, and offering practical advice on when and how to adopt document databases such as SequoiaDB for flexible, high‑performance applications in the big‑data era.
This article describes how MySQL architectures evolve from a single instance to vertical splitting, master‑slave replication, horizontal sharding, and cloud‑based SaaS solutions to handle increasing traffic, data size, and operational requirements while maintaining scalability and availability.
This article summarizes publicly available information and personal experience to outline Google’s product portfolio, design principles, workload categories, and the distinction between its giant and medium‑sized data centers, providing a speculative view of the company’s overall architecture.
The article details how Qihoo 360 adopted MongoDB since 2011, scaling to over 100 applications, 1,500 instances and 20 billion daily queries, and shares their architectural choices, backup strategies, best‑practice recommendations, and advice for teams considering MongoDB in large‑scale, cloud‑native environments.
The article reveals how operating‑tool systems, often underestimated, demand high technical rigor to ensure automation success rates and absolute reliability for emergency actions, requiring sophisticated failure handling, capacity awareness, and scalable design—challenges comparable to core online services.
The article explains MySQL Cluster as a real‑time, ACID‑compliant, in‑memory distributed database, describes its node architecture, transparent sharding for linear scalability, various SQL/NoSQL access methods, and presents a benchmark achieving 200 million queries per second on 32 data nodes.
The article examines how Facebook’s massive data centers operate at scale, detailing server serviceability, automated fault‑diagnosis systems like CYBORG, staffing ratios, and the engineering practices that enable reliable service for billions of users worldwide.
NGINX achieves high performance and scalability by using a multi‑process architecture with a single master process, multiple single‑threaded non‑blocking worker processes, and dedicated cache loader/manager processes, limiting worker count per CPU core and handling many connections via an event‑driven state machine.
This article introduces MySQL Cluster, an in‑memory, real‑time, ACID‑compliant distributed database, explains its multi‑master architecture, data, application and management nodes, transparent partitioning, NoSQL APIs, and presents a benchmark achieving 200 million queries per second on up to 32 nodes.
This article explains how to achieve ultra‑high HTTP request rates by using load‑generating tools such as Apache Bench, httperf, JMeter and Tsung, discusses Linux socket limits, provides step‑by‑step installation and configuration examples for Tsung on CentOS, and shows how to run tests and interpret the results for building a high‑performance web cluster.
This article details how Dianping’s sub‑40‑person operations team structures its groups, designs a dual‑datacenter architecture, and creates comprehensive monitoring, automation, configuration, and analysis systems—including Zabbix, Cat, workflow, Button, and a custom radar platform—to achieve high‑availability, self‑service, and continuous improvement.
The article outlines how narrowing operational permissions, automating processes, and adopting data‑center operating systems enhance security, scalability, visibility, and fault handling while reducing manual effort and enabling rapid, low‑risk changes across large infrastructures.
This article examines WhatsApp’s high‑reliability architecture that supports 450 million users, detailing its Erlang‑based backend, hardware choices, scaling techniques, performance metrics, monitoring tools, and lessons learned from achieving up to two million concurrent connections on a single server.
This article examines key considerations for building highly scalable and available systems on Azure, outlining four architectural dimensions—scalability, availability, manageability, and feasibility—while discussing patterns, anti‑patterns, measurable resources, queue‑based load balancing, authentication services, and common pitfalls such as configuration errors and SQL injection.
This article analyzes common push‑service questions, presents a memory‑leak case from an IoT MQTT middleware, and provides detailed Netty design guidelines—including file‑handle limits, CLOSE_WAIT handling, thread and task management, heartbeat tuning, buffer allocation, memory‑pool usage, logging pitfalls, TCP and JVM optimizations—to help engineers build stable, high‑performance million‑connection push servers.
The article summarizes Google’s recent papers that outline an architectural guide for creating elastic, scalable web applications on Google Cloud Platform, describing key components such as regions, zones, load balancers, autoscalers, Cloud SQL, Cloud Storage, and deployment best‑practices.
This article distills the key takeaways from the book “50 Principles of High‑Scalability Websites,” presenting concise guidelines on avoiding over‑design, planning capacity, simplifying architecture, optimizing DNS and assets, leveraging horizontal scaling, proper database usage, caching, fault isolation, asynchronous messaging, and continuous learning to build robust, easily extensible web systems.
This article examines how WhatsApp grew its user base, message volume, and infrastructure dramatically over two years while keeping the engineering team at ten, detailing the architectural choices, Erlang‑based stack, hardware scaling, decoupling strategies, and the operational challenges that led to both performance gains and a major outage.
The article breaks down Facebook's multi‑layered architecture, explaining how its PHP‑based web tier, Thrift‑driven services, Memcached caching, and specialized Erlang and C++ servers work together to deliver the News Feed and real‑time chat at massive scale.
The article details Facebook’s LAMP‑based architecture, describing how HipHop compiles PHP to C++, Thrift‑based services in PHP, C++, and Java run on custom servers, and how MySQL, Memcached, Cassandra, HBase, Hadoop, Hive, Scribe, BigPipe, Varnish, Haystack and other components together enable handling over 60,000 servers, 300 TB of cached data, 1 trillion daily clicks and petabytes of storage.
This article analyzes common push‑service questions, presents a real‑world IoT case of memory leakage, and outlines key design points—including kernel limits, CLOSE_WAIT handling, heartbeat configuration, buffer management, memory pooling, logging pitfalls, TCP tuning, and JVM settings—to build a scalable, stable Netty‑based push server for millions of concurrent connections.
The article provides a comprehensive overview of Flickr’s web architecture, detailing its massive traffic statistics, load‑balancing, PHP application servers, MySQL dual‑tree sharding, Memcached caching, search engine integration, and the evolution of its replication and scaling strategies.
The article explains how WeChat’s backend handled the massive surge of red‑packet traffic during the 2015 New Year by employing loss‑tolerant services, flexible availability, modular design, asynchronous processing, and overload protection to ensure fast, accurate, and stable user experiences.
This article compiles practical discussions on database sharding and partitioning, covering criteria such as user ID, time, and hash, redundancy techniques, caching with Redis, query handling, pagination, consistency, and distributed transaction approaches for scalable backend systems.
This interview with Alibaba’s DRDS expert Shen Xun reveals how the distributed relational database service evolved from an internal tool to a cloud‑native solution, detailing its architecture, scalability advantages, middleware ecosystem, and real‑world performance during massive events like Double‑11.
The article examines the inherent scalability constraints of the JVM, explains how memory walls and pause times affect enterprise Java applications, and presents practical tuning parameters and architectural considerations to mitigate performance bottlenecks in production environments.
This article details a straightforward architecture built on HAProxy, PHP, Redis, and MySQL that reliably handles one billion weekly requests, explains the underlying hardware, monitoring setup, and provides practical lessons and future scaling strategies for high‑traffic web services.
This article examines Stack Overflow’s highly efficient architecture—using a mix of Linux and Windows, vertical scaling, extensive caching, SSD storage, and minimal hardware—to deliver billions of page views while keeping server utilization remarkably low.
From the evolution of single‑machine file systems to today’s distributed, erasure‑coded cloud storage, this article examines why storage has become increasingly complex, the limitations of traditional replication, and how Qiniu’s next‑gen architecture leverages EC, faster repairs, and cost reductions to meet scalability, reliability, and availability demands.
This article explains what MySQL table partitioning and sharding (splitting tables) are, compares their implementation, data handling, performance impact, and difficulty, and shows how they can complement each other for high‑traffic, large‑data scenarios.
When deciding whether to build or buy a mobile game backend, evaluate the backend’s impact on gameplay, your team’s expertise and resource availability, the long‑term maintenance and update commitments, risk tolerance, and budget constraints to choose the most scalable, cost‑effective solution.
The article traces web architecture’s evolution from simple single‑server Java/JSP sites through memory caching, load‑balanced multi‑machine logic, read/write separation, master‑slave replication or message‑queue syncing, horizontal/vertical sharding, and finally multi‑server web‑server deployments behind load balancers to achieve scalability and reliability.
