This article presents a comprehensive overview of the 58.com alliance advertising platform, detailing its business model, core modules such as SSP, DSP, media and creative platforms, system architecture, performance optimizations, big‑data processing, and future directions for programmatic creative generation.
This article details the overall architecture of JD.com Data Science's JDDLB load balancer, its high‑performance and high‑availability features, and presents a comprehensive performance comparison of SSL/TLS offloading using Intel QAT acceleration cards, including async processing, user‑space driver zero‑copy implementation, crash analysis, and process‑level engine scheduling.
This article presents Didi's three‑year experience with Presto, detailing its architecture, low‑latency design, large‑scale deployment, extensive Hive compatibility work, resource isolation, Druid connector integration, usability enhancements, stability engineering, performance tuning, and future directions for the ad‑hoc query engine.
This article details a real‑world investigation of Spark SQL job latency on a YARN cluster, explains how switching the scheduler to FAIR mode, creating resource pools, and consolidating small Parquet files dramatically reduced scheduler delay and cut execution time from over 100 seconds to under 20 seconds.
This article explains how to design, tune, and optimize an Elasticsearch‑based data platform handling hundreds of billions of records, covering Lucene fundamentals, shard routing, index and query performance tricks, and practical benchmark results for large‑scale deployments.
Optimizing app launch on iOS and Android—through deep‑link handling, H5 splash strategies, On‑Demand Resources, custom WKWebView schemes, reduced download size, static and runtime .so loading flags, thread‑pool tuning, method swizzling, actor‑based concurrency, and using Instruments, MetricKit and Android Profiler—prevents user abandonment and boosts conversion.
This article analyzes the challenges of traditional monolithic big‑data architectures, explains how Tencent Cloud EMR integrates Alluxio for compute‑storage separation, presents detailed performance benchmarks showing 20‑50% bandwidth reduction and 5‑40% query speedup, and outlines the specific tuning measures applied.
This article details how the DTCC2020 guest speaker from Tongcheng‑Elong introduced BaikalDB’s distributed columnar storage, covering internal and external motivations, technology comparison, architecture, implementation tricks, performance gains in production, and future hybrid row‑column research directions.
This article examines JD Intelligent Cloud’s self‑developed modular servers, highlighting their innovative front‑back I/O design, liquid‑cooling integration, custom Intel CPUs, and performance gains that address high PUE, deployment complexity, and scalability challenges in modern data centers.
Learn how to accelerate Python image preprocessing by leveraging the built‑in concurrent.futures module to run tasks across all CPU cores, turning a single‑core script that takes seconds into a multi‑core version that finishes in under two seconds with only three extra lines of code.
This article examines the pervasive small‑file issue in big‑data environments, explains its impact on storage and processing performance, and presents a comprehensive set of solutions—including file merging, Hadoop archives, SequenceFiles, HBase, CombineFileInputFormat, and Spark/Flink strategies—to mitigate metadata overhead and improve I/O efficiency.
This article describes how Zhihu’s Moneta application migrated billions of rows of user‑read data to the open‑source MySQL‑compatible NewSQL database TiDB, detailing the architectural redesign, performance improvements, migration challenges, and future expectations for TiDB 3.0.
This article details how eBay's ADI Hadoop team tackled a massive 10 PB, 10‑million‑file migration by optimizing DistCp with Fastcopy, load‑balancing, ACL handling, and failure recovery, ultimately completing the transfer within a two‑hour window while preserving cluster stability and performance.
This article analyzes common database performance bottlenecks, explains when to apply IO‑ or CPU‑focused optimizations, and provides a detailed guide to horizontal and vertical sharding, partitioning strategies, tooling choices, implementation steps, and practical pitfalls for scalable systems.
This guide walks through installing GraalVM, adding the native‑image tool, creating a Spring Boot 2.4 project, configuring it for GraalVM, building a native executable, and comparing startup time and memory usage, showing how GraalVM can reduce startup from over a second to 0.2 seconds and memory from 480 MB to 25 MB.
Suning’s senior architect Yang Zhaohui explains how his team rebuilt the tag platform with ClickHouse, using RoaringBitmap and custom optimizations to achieve second-level queries on billions of user tags, dramatically cutting response time, reducing hardware costs, and enabling real-time marketing insights.
This article details how Alibaba Cloud transformed RocketMQ with a Kubernetes‑based operator, optimized message filtering by indexing MessageType for up to 30% CPU reduction, and introduced a POP consumption model that eliminates rebalance delays, achieving stable performance during the 2020 Double‑11 peak.
The article details how Vivo’s official mall migrated from a heavy SPA to a Nuxt‑based server‑side rendering solution, employing LRU page caching, component and API caches, concurrent requests, selective client‑only rendering, robust degradation tactics, automated CI/CD deployment, and monitoring, which together boosted QPS from 125 to 6 000—a roughly fifty‑fold performance gain and markedly better user experience.
This article explains how Alibaba’s Serverless Cloud Development Platform, built on Function Compute, enables thousands of engineers to develop, deliver, and operate large‑scale applications faster and more reliably by providing integrated solutions, customizable capabilities, and comprehensive stability and operations tools.
The article provides a comprehensive overview of the Lustre parallel file system architecture, its core components, POSIX compliance, scalability, high‑performance networking, security features, data layout mechanisms, and performance considerations for large and small files, along with practical optimization tips for HPC environments.
This article details Industrial Bank's migration to Dubbo micro‑services, the performance and high‑availability challenges of managing over 20,000 services with Zookeeper, and the concrete optimizations—delayed subscription, multiple‑registry mode, and per‑node registration—that enabled stable, large‑scale service discovery.
This article documents a step‑by‑step MySQL data migration case involving 140 million rows, covering export via SELECT INTO OUTFILE, import with LOAD DATA INFILE, and a series of performance tweaks such as index removal, session variable tuning, engine selection, and monitoring to achieve sub‑10‑minute load times.
This article explains how Spark can be combined with MongoDB for large‑scale data processing, covering Spark fundamentals, comparisons with HDFS, practical integration patterns, performance benefits, real‑world case studies, and detailed code examples for deployment and analytics.
This article explains how to design a flash‑sale (seckill) system that handles massive concurrent reads and writes by applying principles such as minimizing data and request volume, shortening request paths, eliminating single points of failure, and employing layered caching, traffic shaping, and robust high‑availability strategies.
Cache leverages the space‑time trade‑off to boost performance, but introduces complexity; this guide explains its fundamental idea, local vs distributed solutions, popular Java caching libraries, and key read/write patterns such as Cache‑Aside, Read‑Through, and Write‑Behind, highlighting benefits, drawbacks, and best‑practice considerations.
This article explains why multithreading improves program performance, distinguishes latency reduction from throughput increase, and provides practical formulas to determine the ideal number of threads for both I/O‑bound and CPU‑bound workloads, helping you fully utilize hardware resources.
The analysis reveals that on Vivo devices running Android 9, a static AsyncTask SERIAL_EXECUTOR blocks the UpdateBottomFlagTask in several order‑related activities, leaking the Activity context, and the fix replaces this executor via reflection with a custom parallel SerialExecutorProxy to prevent the leak.
This article compares Apache Kylin and ClickHouse, two popular big‑data OLAP engines, by examining their technical principles, storage structures, optimization techniques, and ideal application scenarios to help readers make an informed technology selection.
The Xianyu team optimized their Flutter‑based app by creating a lag‑diagnosis tool, minimizing unnecessary widget rebuilds, deferring load‑more updates, using low‑resolution placeholders, prefetching data with skeleton screens, reusing FlutterBoost containers and early data passing, which together cut jank by half, shaved ~300 ms off page loads and raised average FPS by about three points on low‑end devices.
The article examines the hidden lock‑contention issue in InnoDB’s Adaptive Hash Index (AHI) observed during high‑concurrency sysbench runs, explains why multiple threads block on the AHI hash table’s exclusive lock, and presents a lightweight fallback optimization that checks the lock before building AHI, reducing lock wait time and stabilizing QPS performance.
This article details the background, system design, and successive optimizations of iQIYI's mobile Application Performance Monitoring (APM) network monitoring solution, covering SDK and backend architecture, DNS and weak‑network enhancements, gateway strategies, retry mechanisms, and the resulting significant reduction in error rates.
This article explains how Redis achieves high performance by storing data in memory, using optimized data structures like SDS, linked lists, ziplists, hash tables, and skip lists, and employing an I/O multiplexing single‑threaded reactor model to minimize latency and overhead.
This article outlines essential database design concepts, covering the relationship between source documents and entities, primary‑foreign key design, basic table characteristics, normalization standards, handling many‑to‑many relationships, primary key strategies, data redundancy, ER‑diagram considerations, view usage, and performance‑boosting techniques.
This article explains how JD Intelligent Cloud’s bare metal cloud hosts delivered near‑physical performance, automated operations, and strong isolation to support the 11.11 shopping festival’s billions of visits and millions of concurrent requests, detailing architecture, performance gains, and integration with Kubernetes.
Meituan’s sales system “Qingtian” boosted OLAP performance by migrating Apache Kylin’s build engine from MapReduce to Spark, consolidating Hive files, refining dictionary creation, applying a By‑layer algorithm, and bulk‑loading cuboid files to HBase, cutting resource consumption and halving build time, ultimately reaching a 100 % SLA.
The 2017‑initiated vivo distributed tracing system leverages a JavaAgent‑based micro‑kernel architecture, using ByteBuddy for non‑intrusive bytecode instrumentation, a Disruptor lock‑free queue, and Kafka to capture Trace/Span data—including cross‑thread propagation—while employing sampling, degradation, and JVM metrics to ensure 94% adoption stability.
In 2020 Alibaba upgraded its Double 11 e‑commerce venue by introducing a low‑risk data‑driven Server‑Side Rendering (SSR) solution that lifted the page‑open rate to 82.6% and increased UV click‑through by up to 5%, while detailing the architectural changes, performance metrics, implementation challenges, and measurable business impact.
NetEase’s XRN platform boosts Web and React Native performance by eliminating duplicate bundles, pruning unused code, importing only needed polyfills, using dynamic imports and bundle splitting, pre‑loading assets, lazily loading images and lists, limiting first‑screen rendering, and applying memoization, raising Lighthouse scores from 50.3 to 80.4 and cutting RN load time by 40 %.
This article explains how to improve the performance of Java code that compresses multiple images into a zip archive by replacing unbuffered streams with buffered I/O, then using NIO channels, direct buffers, memory‑mapped files, and pipes, achieving a reduction from 30 seconds to about 1 second.
This article demonstrates how to dramatically reduce the time required to compress multiple large images in Java by progressively applying buffered streams, NIO channels, direct buffers, memory‑mapped files and pipe techniques, measuring each optimization and explaining the underlying I/O mechanisms.
This article introduces ClickHouse, compares column‑store and row‑store databases, shows how migrating a 50‑million‑row MySQL table to ClickHouse reduced query time from minutes to under one second, and shares practical installation, migration, performance testing, and synchronization tips.
This article introduces ClickHouse, compares column‑ and row‑oriented storage, explains a real‑world migration from MySQL to ClickHouse that reduced a 3‑minute query to under one second, details installation, migration methods, performance results, synchronization options, and common pitfalls.
Ctrip’s hotel data team tackled a 3 TB daily data load by building a ClickHouse cluster on VMware, creating custom sync and execution tools, applying query optimizations, and handling merge and memory errors, ultimately achieving over 400% performance gains across multiple reporting themes.
The article details Xiaomi's growth analytics platform evolution from a Lambda architecture using SparkSQL, Kudu, and HDFS to a streamlined MPP solution with Apache Doris, covering performance gains, real‑time data ingestion, query tuning, and operational improvements for large‑scale analytics.
During the pandemic, the Yunnan “Anti‑Epidemic” mini‑program faced over 80 W requests per minute, prompting an analysis of how cloud development’s serverless, auto‑scaling architecture—covering data pipelines, cloud functions, and cloud databases—can ensure high performance, availability, and cost‑effective handling of massive concurrent traffic.
Understanding CPU cache hierarchy, its speed advantages over memory, and the mechanics of cache lines, tags, and offsets reveals why code that maximizes cache hit rates—through sequential data access, branch prediction, and core affinity—can run dramatically faster on modern processors.
The iQIYI mobile APM system provides real‑time, user‑level network monitoring with classified error detection, cloud‑controlled SDK sampling, second‑level backend storage, and web dashboards, while employing DNS three‑layer caching, weak‑network grading, gateway multiplexing, super‑pipeline proxies and layered retry strategies, reducing Android error rates from 5.3 % to 0.48 % and iOS from 4.63 % to 0.35 %.
This article demonstrates how to replace complex SQL search queries with a Redis‑based solution by caching intermediate result sets using sets and sorted sets, optimizing performance through multi‑command transactions, and adding pagination, offering a practical backend development pattern for high‑traffic e‑commerce search.
This article describes the architecture of JD.com’s JDDLB public‑traffic load balancer and details how offloading CPU‑intensive SSL/TLS cryptographic operations to Freescale C291 acceleration cards—via custom NGINX modules, OpenSSL Engine integration, and synchronous/asynchronous driver interfaces—significantly improves connection‑establishment rates and overall throughput.
This article explains how to replace a 29‑day serial processing job for 100,000 string records with a Redis‑backed asynchronous task queue powered by Node.js, PM2 clustering, and distributed locking, achieving a ten‑second runtime for 20 sample tasks.
This article details how Instagram migrated its massive Python 2/Django codebase to Python 3, describing the motivations, technical obstacles such as Unicode, pickling, iterator changes, the tools and strategies used, and the resulting CPU, memory, and latency improvements across billions of users.
This article shares a front‑end engineer’s practical experience optimizing Vue SSR for a rapidly growing medical platform, covering background, SSR pros and cons, conventional and deep‑level caching strategies, CDN integration, network tuning, and the trade‑offs of each approach.
This article presents a comprehensive overview of NetEase's use of Impala as an OLAP query engine, detailing its architectural advantages, performance benefits, enhancements such as management servers, metadata synchronization, high‑availability via Zookeeper, expanded storage support, and real‑world deployment cases in the "Mammoth" platform and NetEase Cloud Music.
By parallelizing Spark’s driver‑side commit, trash, and move phases—previously single‑threaded operations that caused costly copy‑on‑rename when writing massive files to object storage—the Tencent Cloud EMR case achieved over a tenfold (1,100 %) speedup, making object storage a viable alternative to HDFS.
This article introduces Tencent's self‑built Angel distributed machine‑learning platform, describes its architecture and deep‑learning extensions (Parameter Server and AllReduce), explains how it powers the advertising recommendation pipeline with models such as DSSM, VLAD and YOLO, and presents extensive training‑level optimizations that yield multi‑fold performance improvements.
Didi’s three‑year journey with Presto transformed it into the company’s primary ad‑hoc and Hive‑SQL acceleration engine, serving over 6 000 users, processing 2‑3 PB of HDFS data daily, and achieving major gains in stability, performance, cost, and usability through extensive architectural tweaks, resource isolation, connector extensions, and monitoring enhancements.
This article provides a comprehensive introduction to Elasticsearch, covering data types, the role of Lucene, cluster architecture, node roles, discovery mechanisms, shard and replica management, mapping, installation, health monitoring, indexing workflow, storage internals, refresh and translog processes, segment merging, and practical performance and JVM tuning tips.
This article explains how to pick the most suitable cloud storage—cloud disk, file service, or object storage—by using four easy-to-remember mnemonics that cover functionality, performance, concurrency, and latency, helping readers avoid complex parameter analysis.
Facing growing log volumes and high Elasticsearch costs, Ctrip migrated its 200 TB daily logs to ClickHouse, achieving up to 38× faster queries, 60 % lower server resources, and simplified operations through columnar storage, sharding, and custom dashboards, while detailing deployment, tuning, and common pitfalls.
This article explains the classification of OLAP data warehouses by data volume and modeling approach, compares MOLAP, ROLAP, HOLAP and HTAP, reviews popular open‑source ROLAP systems, and details advanced performance‑boosting techniques such as MPP architectures, cost‑based optimization, vectorized execution, dynamic code generation, and runtime filtering.
This guide outlines practical capacity‑planning, index‑design, and write‑performance tuning for Tencent Cloud Elasticsearch clusters, covering compute and storage sizing, optimal shard counts, rollover strategies, bulk API settings, health monitoring, and common troubleshooting steps to ensure stable, high‑throughput search services.
This article walks through creating a Web‑based 3D endless‑runner game using Oasis Editor, covering scene design, asset preparation, scripting, shader development, performance tuning, and the migration from Oasis 3D V2 to V3, with practical code examples and visual illustrations.
This article examines common pitfalls in code quality, testing, and development efficiency, analyzes their root causes, and proposes concrete, actionable solutions such as performance optimization, stricter coding standards, regular code reviews, improved testing practices, and enhanced team collaboration to boost overall software project success.
This article examines Alibaba Entertainment's three‑stage evolution of OTT front‑end technology—from early WebView implementations through a custom Blitz engine to the current Weex (RAX) solution—detailing implementation steps, performance‑boosting techniques, and measurable results that cut page load times by over 50%.
This article introduces ClickHouse, a column‑store OLAP database, explains its advantages over row‑store systems, guides installation on macOS, details migration strategies from MySQL, presents performance benchmarks showing 200‑fold speedups, and discusses synchronization methods, pitfalls, and practical tips.
This article explains how to construct a reliable, high‑throughput microservice platform using open‑source components such as Consul, an access layer, and token‑based multi‑tenant service discovery, covering architecture, registration, heartbeat handling, performance optimizations, high‑availability strategies, graceful shutdown, and deployment best practices.
This article introduces ClickHouse as a high‑performance columnar OLAP database, compares it with row‑based systems, details migration methods from MySQL, presents performance benchmarks showing dramatic speed gains, discusses data synchronization strategies, explains why ClickHouse is fast, and shares common pitfalls and solutions.
This article details how JD Daojia's order center migrated from a simple MySQL‑backed system to a high‑throughput Elasticsearch cluster, describing each architectural phase, performance tuning measures, dual‑cluster real‑time backup, version upgrades, data synchronization strategies, and the key pitfalls encountered such as deep pagination and FieldData memory issues.
This article details how JD.com's order center migrated its Elasticsearch cluster from a simple, default‑configured setup to a highly available, multi‑replica, dual‑cluster architecture with version upgrades, data synchronization strategies, and performance optimizations to support billions of documents and hundreds of millions of daily queries.
NetEase Cloud Music rebuilt its React Native stack from version 0.33 to 0.6, introducing automated one‑click deployment, a custom codemod that migrated all apps in a day, and a three‑platform toolchain that unified iOS, Android, and web, boosting checkout reach to 99% and scaling to over ten production RN applications.
The article details how Didi’s large‑scale Ceph deployment suffered from high tail latency due to long‑held and coarse‑grained locks, and describes a series of fixes—including asynchronous read threads, fine‑grained object caches, per‑thread lock‑free logging, and lock‑free filestore apply—that cut latency by up to 90 % and more than doubled read throughput.
Elasticsearch, built on Apache Lucene, provides a distributed, near‑real‑time search platform that scales to billions of documents across thousands of nodes, supporting use cases such as log analytics, time‑series monitoring, and product search, while Tencent’s CES adds advanced availability, performance, and cost‑optimizing features.
This article examines the unique memory‑leak challenges of desktop front‑end applications, explains JavaScript memory management, presents analysis methods using Chrome's Memory panel and a binary‑search approach, and offers practical optimization, verification, and monitoring techniques to keep memory usage low during rapid feature iterations.
This article walks through a systematic performance audit of the Jingxi Mini‑Program, identifies seven key penalty items—including excessive WXML nodes, oversized images, tiny tap targets, redundant network calls, and unbound data—and details concrete optimizations that reduced the audit score to a perfect 100.
This article explains the purpose and practical benefits of bitwise operations in Java, describes signed number representations (sign‑magnitude, ones' complement, two's complement), shows how bits can optimize time and space, and details the seven common bitwise operators with examples and a reference table.
Tencent Cloud MySQL 8.0 builds on official MySQL improvements with eight enterprise‑grade enhancements—TDE encryption, kernel‑level audit, multi‑queue thread pool, strong consistency, AEP SSD support, lightweight AP mode, hotspot updates, and SQL rate‑limiting—delivering high‑performance, secure, and scalable solutions for e‑commerce, gaming, finance, and new‑retail workloads while paving the way for cloud‑native, AI‑driven database evolution.
This article explains common I/O and CPU bottlenecks in databases, compares horizontal and vertical sharding techniques, outlines practical partitioning strategies, introduces popular sharding tools, and provides step‑by‑step guidance for implementing and scaling sharded architectures.
This article examines common Elasticsearch interview questions, explains why initial searches can be slow, and provides practical strategies such as leveraging filesystem cache, data pre‑heating, cold‑hot index separation, minimal document design, and scroll or search_after APIs to dramatically improve search performance and pagination efficiency.
Gradle 6.6 introduces experimental configuration caching, runtime classpath normalization, and a new credentials API, all aimed at dramatically speeding up builds, improving cache hit rates, and simplifying secure artifact publishing for Java projects.
Faced with tens of thousands of dirty records and no bulk‑delete API, the author iteratively optimized a deletion script—starting with simple serial requests, then adding asynchronous HTTP calls, and finally employing multithreading—to achieve rapid, reliable removal of hundreds of entries per second.
This article details the comprehensive redesign of JD VSP's PC platform, covering the motivations for the overhaul, the adoption of Vue, TypeScript, Gaea scaffolding, Legao component library, performance enhancements with icon fonts and PWA caching, and the implementation of robust development tooling such as ESLint and code‑generation workflows.
By establishing measurable standards, profiling the launch flow, and applying task‑governance techniques—delaying, shrinking and splitting blocking tasks—plus home‑page shortcuts like ad pre‑fetch, background view creation and SharedPreferences caching, Xianyu’s Android cold‑start on low‑end phones dropped from about ten seconds to under five, with the first screen in ~1.3 seconds and full launch in 4.5 seconds.
This article explains the core principles of performance optimization, distinguishes active and passive approaches, and provides detailed MySQL passive tuning techniques—including index usage, data partitioning, slow‑query logging, and read/write splitting—to prevent and resolve slow SQL queries.
This article explains how Bilibili's GOIM live chat system was architected and continuously optimized to achieve high stability, high availability, and sub‑second latency through modular backend components, memory and module optimizations, and multi‑IDC network improvements.
The article describes how a live‑room video‑gift feature originally built for mobile was ported to a web client by extracting separate color and alpha video streams, compositing them on canvas, then migrating the per‑pixel blending to WebGL shaders, which cut CPU usage dramatically, raise frame rates to about 60 FPS, and outline further optimisations such as pre‑loading, mobile support, and possible MSE or WebAssembly approaches.
This article provides a comprehensive overview of large‑scale website architecture, covering characteristic traits, performance and availability goals, layered design patterns, security measures, scalability and extensibility strategies, evolution stages, capacity estimation, and practical optimization techniques for e‑commerce platforms.
When a business platform’s settings page took 36 seconds to load, we used Chrome’s Network timing, Python profiling, and MySQL query analysis to identify backend bottlenecks, then applied redesign, thread‑pooling, batch queries, and ORM optimizations, reducing response time from 37.6 s to 1.47 s.
The article details iQIYI’s migration from a Hive + MySQL OLAP stack to Apache Kylin, describing the system’s architecture, typical use cases, performance gains, independent HBase deployment, service platform for monitoring, and future plans such as automated cube building and clustering.
This article traces the evolution of database autonomy from manual, knowledge‑driven operations through tool‑assisted and expert‑level stages to cloud‑native intelligent services, and details Tencent's DBbrain platform, its architecture, performance‑optimization, security, monitoring, cost‑based analysis, and future self‑healing capabilities.
This article details JD’s 京喜 mini‑program performance overhaul, covering user‑centric metrics, dual‑thread architecture, code‑size reduction, asset optimization, plugin‑based rendering, data‑handling strategies, and memory management to dramatically improve startup, rendering, and overall user experience.
Apache Spark 3.0 introduces a suite of performance enhancements, richer APIs, improved monitoring, SQL compatibility, new data sources, and ecosystem extensions, including Adaptive Query Execution, Dynamic Partition Pruning, Join Hints, pandas UDF improvements, and accelerator‑aware scheduling, to boost scalability and ease of use for big‑data workloads.
ClickHouse, the high‑performance columnar OLAP DBMS behind Yandex.Metrica, combines complete DBMS capabilities, column‑oriented storage with compression, vectorized execution, flexible table engines, multi‑master clustering, and extensive SQL support, offering fast online queries and scalable distributed processing for massive data workloads.
Youku’s player was rebuilt with a pipeline‑pattern architecture that unifies playback stages into a modular framework, enabling plug‑in customization, multi‑source support, reduced memory and startup overhead, clearer structure, and improved stability and extensibility for future business innovations.
This article explains the significance of data compression, outlines when it should be applied, compares lossless algorithms, discusses segment selection, and details how Kafka and RocketMQ implement message compression to improve throughput while balancing CPU, storage, and network resources.
This article explains the principles of MySQL performance optimization, distinguishes active and passive approaches, and provides concrete solutions for slow single queries, partially slow queries, and overall slow queries through proper indexing, data partitioning, slow‑query‑log configuration, and read‑write splitting.
DiDi’s IFX AI inference platform, built since 2018, uses a four‑layer architecture spanning access, software, engine, and compute to deliver cloud, edge, and device inference with high‑performance kernel optimizations, model and binary compression, uniform multi‑framework deployment, automated testing, and end‑to‑end security for billions of daily calls.
This article explains the principles and practical implementation of the Java off‑heap cache framework OHC, detailing its architecture, memory allocation, serialization, configuration, and real‑world performance results within the MaFengWo recommendation engine, illustrating how it reduces latency and improves cache hit rates.
This article presents a comprehensive guide to database design, covering the relationship between source documents and entities, primary and foreign keys, table properties, normalization standards, handling many‑to‑many relationships, key selection, data redundancy, ER diagram considerations, view usage, intermediate and temporary tables, integrity constraints, the "Three‑Few" principle, and practical techniques for improving database performance.
Tencent Meeting migrated its quality‑analysis system to Tencent Cloud Elasticsearch, tackling OOM failures, 3 M/s write spikes and scaling limits by adding multi‑AZ deployment, leaky‑bucket rate limiting, streaming aggregation checks, optimized merge and translog handling, plus hot‑warm storage, ILM, multi‑disk and off‑heap caching, cutting cluster size from 15 000 to under 300 nodes while maintaining high availability and performance.
To dramatically improve Hive UDF performance when converting coordinates to administrative districts, this article compares two implementation strategies, details the technical challenges of repeatedly loading a 157 MB Geo data file, and presents a static‑cached solution that reduces query time from seconds to milliseconds, achieving roughly a 140‑fold speed increase.
