This article details the business background, technical challenges, and the step‑by‑step implementation of an intelligent data warehouse on Alibaba Cloud MaxCompute, covering offline data pipelines, metric calculation, data analysis, and future plans for data lake and AI‑driven analytics.
The Apache Hudi Asia technical salon held in Beijing on March 29 gathered over 230 on‑site participants and 16,000 online viewers, featuring expert talks from leading Chinese tech companies that showcased real‑world Hudi implementations, performance optimizations, and future roadmap for data‑lake technologies.
The first Apache Hudi Asia Summit in Beijing attracted over 230 attendees, featuring technical discussions on data lake optimization and case studies from companies like Fastly and Meituan.
The article explains how data lakes serve as a foundational component for observability by aggregating raw, diverse data for advanced analysis, while also outlining the technical, cost, and scalability challenges that make them unsuitable for every organization.
This article presents a comprehensive exploration of using Apache Paimon and Flink to design lake tables that support minute‑level latency, low cost, and unified batch‑stream processing for advertising data, covering schema design, partitioning strategies, performance trade‑offs, cost analysis, and operational best practices.
Alibaba Mama’s advertising platform migrated from a monolithic Flink‑Kafka pipeline to a layered Paimon lakehouse, adding DWS upsert support and multi‑layer storage, which delivers minute‑level data freshness, cuts latency by 2.5 hours, reduces resource use over 40 %, halves development effort and achieves ≥99.9 % availability.
This article examines how Apache Iceberg’s snapshot‑based ACID transactions, logical‑physical partition evolution, and COW/MOR update modes enable efficient real‑time data lake ingestion, and demonstrates AutoMQ’s Kafka‑to‑Iceberg Table Topic solution that simplifies schema management, reduces latency, and cuts operational costs.
This article introduces DeepSeek Smallpond, a lightweight yet high‑performance AI data‑processing engine built on Ray and DuckDB, explains its dual Dataframe and LogicalPlan APIs, showcases integration with Volcano Engine's AI Data Lake LAS, and provides practical code examples for distributed processing, multimodal storage, and RAG pipelines.
This article presents Douyin Group’s ByteLake, a heavily customized Apache Hudi‑based data lake table framework, detailing its core concepts, metadata services, write and read optimizations, operational challenges, a fully managed table management service, and its integration with the Amoro open‑source platform.
Midea Building Technology shares how its IoT data platform leverages Alibaba Cloud EMR Serverless Spark, Hudi Lakehouse, and Serverless StarRocks to achieve real‑time ingestion, massive scale processing, AI‑driven analytics, and significant performance and cost improvements for building‑system management.
This article provides a comprehensive guide to designing and implementing a data middle platform, covering its core essence as a business-data fusion mechanism, six architectural modules from ingestion to performance optimization, three implementation phases from requirements to applications, and key pitfalls like business-technology misalignment and over-engineering.
This article presents JD Advertising's engineering experience with Apache Doris, describing the evolution from a data‑lake cold‑data solution to a native cold‑hot tiering approach, detailing performance regressions after upgrading to Doris 2.0, and outlining a series of optimizations for query speed, CPU and memory usage, schema‑change efficiency, and automated data migration and restoration.
This article traces the evolution of data integration from point-to-point and ETL to modern lakehouse architectures, detailing seven key shifts—tool-to-value orientation, downstream-to-upstream integration, manual-to-automated pipelines, professional-to-citizen access, siloed-to-ecosystem scope, batch-to-real-time processing, and isolated-to-unified storage—driven by data elementization trends.
Alibaba Cloud DataWorks’ Data Integration platform, built on Flink CDC, offers a comprehensive, serverless solution for real‑time and batch data lake ingestion, detailing its architecture, elastic scaling, productized use cases, and future roadmap, including AI‑driven diagnostics and expanded source support.
In 2024, JD Retail Technology rolled out a series of breakthroughs—including a major JD APP redesign, data‑driven inventory algorithms, an AIGC content platform, a low‑code national‑subsidy system, a high‑performance data lake, cross‑platform Taro on Harmony, AI‑powered merchant assistants, and immersive XR shopping—showcasing how AI and advanced engineering drive faster fulfillment, richer user experiences, and scalable innovation.
In 2024 JD Retail Technology delivered a series of breakthroughs—including a major JD APP redesign, a data‑driven inventory selection and allocation algorithm that cut stockouts, an AIGC platform for marketing content, a low‑code national‑subsidy system, a large‑scale Apache Hudi data lake, the Taro‑on‑Harmony cross‑platform framework, immersive XR shopping experiences, and a domestic‑chip AI engine—showcasing how advanced AI, cloud, and operations engineering are reshaping e‑commerce.
This article summarizes a talk by Alibaba Cloud senior engineer and Flink Committer Luo Yuxia on the challenges of separating lake and stream storage, introduces the Fluss lake‑stream unified architecture, explains its technical benefits such as second‑level data freshness, unified metadata, efficient changelog generation, and outlines future plans for broader ecosystem integration.
This tutorial walks you through using Alibaba Cloud EMR Serverless StarRocks and Apache Paimon to create a cost‑effective, high‑performance game player profiling and behavior analysis platform, covering data import, materialized view creation, DWD/ADS layer construction, and lakehouse integration.
Tencent's advertising team replaced a traditional HDFS‑Hive warehouse with an Apache Iceberg‑based data lake, adding primary‑key tables, multi‑stream merging, adaptive compaction, and Spark SPJ optimizations to achieve minute‑level feature update latency, 10× back‑fill speed, and up to 60% storage savings.
The 8th Tongcheng Travel Big Data Technology Salon in Suzhou featured four expert talks covering Tencent Cloud’s Meson Spark engine, near‑line computing for travel itineraries, a Flink‑based real‑time risk control system, and Apache Paimon’s latest lake‑warehouse innovations, followed by a data‑driven business perspective session.
Bilibili migrated its massive user‑behavior, commercial AI training, and database synchronization pipelines from Hive and Kafka to an Iceberg‑based streaming‑batch architecture, using Flink and the Magnus optimizer to achieve minute‑level freshness, reduce CPU and memory usage by about 20‑22 %, save roughly 3.55 M CNY annually, and dramatically improve query latency and join performance.
This article presents Bilibili's end‑to‑end exploration of a streaming‑batch unified data pipeline built on Apache Iceberg, detailing the original and iterated architectures for massive user behavior transmission, online AI training, DB synchronization, and dimension‑join, along with performance gains, cost savings, and future plans.
iQIYI integrates Alluxio with its QBFS multi‑AZ unified scheduling system, automatically caching hot tables, applying table‑level policies, page‑level storage and AZ‑aware worker selection, which together cut cross‑zone traffic, halve query latency, achieve up to 20× I/O speedup and a three‑fold overall performance boost.
In a panel discussion, experts explain how data‑lake‑warehouse integration, columnar formats like Apache Iceberg, and emerging variant types enable efficient feature engineering, support large‑language‑model workloads, and provide flexible vector storage, thereby driving the evolution of AI from traditional ML to the GenAI era.
This article analyzes the evolution of data‑lake storage acceleration, compares traditional parallel file systems, object‑storage‑based solutions and modern cache‑enabled architectures, and explains how cloud‑native data lakes address scalability, cost, and performance challenges for AI and big‑data workloads.
The article examines the evolution of data lake storage acceleration, compares various solutions, and explains how metadata, read/write, and end‑to‑end optimizations enable scalable, cost‑effective AI and big‑data workloads in cloud‑native environments.
Experts from Kuaishou, former Tencent, Ping An Insurance and others discuss data lake maturity, column‑level governance, resource management of unstructured data, and automated optimization techniques such as Iceberg small‑file merging, highlighting how these advances improve data quality and business decision‑making.
In a panel discussion, experts from Kuaishou, Ping An, and Datastrato explain how data lake architectures, columnar storage formats like Apache Iceberg, and vector‑enabled lake formats are enhancing feature management, supporting generative AI workloads, and accelerating machine‑learning pipelines.
Baidu’s Data Lake Storage Acceleration 2.0 replaces traditional HDFS with a scalable object‑storage foundation, introducing an adaptive hierarchical namespace, high‑throughput streaming engine, RapidFS caching, and fully compatible BOS‑HDFS APIs, thereby delivering up to 70 % higher throughput, lower costs, and seamless migration for big‑data and AI workloads.
Data lakes have evolved from HDFS to object storage, addressing resource inefficiency, scalability limits, and operational burdens; Baidu’s Data Lake Storage Acceleration 2.0 introduces hierarchical Namespace 2.0, a streaming storage engine, RapidFS caching, and a fully HDFS‑compatible BOS‑HDFS layer to boost performance and support massive AI workloads.
Magnus is Bilibili’s self‑developed intelligent service that continuously optimizes Iceberg tables by scheduling snapshot expiration, orphan‑file cleanup, manifest rewriting, and multi‑dimensional data optimizations—including small‑file merging, sorting, distribution, and index creation—while automatically recommending configurations from real‑time query logs, delivering over 99.9% task success and up to 30% scan‑data reduction.
The talk outlines Baidu Smart Cloud's comprehensive cloud‑native redesign—including ultra‑elastic compute, AI‑focused storage, high‑performance networking, AI‑driven operations, and edge‑distributed services—illustrated with automotive and fintech case studies that demonstrate how enterprises can accelerate digital transformation in the AI‑native age.
Shopee’s deployment of StarRocks across DataService, DataGo, and DataStudio demonstrates that its vectorized engine, cost‑based optimizer, and materialized‑view caching can query Hive, Iceberg, Delta Lake and Hudi up to 20,000× faster than Presto, cutting CPU usage and delivering consistently lower latency for complex analytics.
This article presents JD Retail's data lake architecture overhaul, detailing the shortcomings of the Lambda model, the migration to Flink‑Hudi‑Spark pipelines, performance gains, storage savings, unified APIs, and upcoming improvements for resilience and automation.
Amid growing data demands, this article explains the data lake technology maturity curve, detailing lake‑warehouse architectural patterns, design principles, core functionalities, and the four leading open‑source solutions (Hudi, Iceberg, Delta Lake, Paimon) to guide enterprises in building flexible, scalable, and governed data platforms.
The article explains Apache Hudi’s three‑layer architecture and details four major 1.0 enhancements—LSM‑tree timeline, non‑blocking concurrency control, file‑group reader/writer APIs, and function indexes—while providing a brief review and links to the Hudi 1.x RFC.
This article explains the data lake technology maturity curve, covering lake‑warehouse architecture patterns, design principles, core capabilities of major open‑source lake engines (Hudi, Iceberg, Delta Lake, Paimon), and practical application scenarios for modern data‑driven enterprises.
At the 2024 Cloud Xi Conference, Alibaba Cloud unveiled a suite of vectorized big‑data solutions—including the Flash engine for Flink, EMR Serverless Spark with a 300% speed boost, upgraded lakehouse architecture, and real‑world case studies—showcasing massive performance gains, cost reductions, and broader serverless adoption.
This article explains Apache Hudi's incremental processing capabilities, covering an overview of the medallion architecture, detailed configuration for incremental queries, the introduction of Change Data Capture (CDC) with required table properties, and a review of how these features enable richer data insights in modern data lake environments.
This article explains the rapid growth of data-driven businesses, the challenges of traditional data warehouses, and how modern data lake technologies such as Delta Lake, Hudi, Iceberg, and Paimon form a maturity curve that guides enterprises in architecture choices, design principles, core capabilities, and practical applications.
This article provides a comprehensive overview of concurrency control in Apache Hudi, explaining ACID properties, the role of MVCC and OCC, and how Hudi coordinates multiple writers and table services to achieve serializable scheduling while maintaining high performance.
This article explores the evolution of data lakes for AI, discusses the challenges of AI-era data management, introduces Apache Iceberg and its architecture, demonstrates PyIceberg-based AI training and inference pipelines, and presents vector table designs with LSH indexing and performance optimizations.
The article details Kuaishou's adoption of a data lake powered by Apache Hudi, covering the challenges of growing data warehouses, the migration from Hive to Hudi, concrete business case studies, promotion strategies, and key takeaways for large‑scale data engineering.
This article examines the challenges of Hudi metadata stored on HDFS, introduces the independently developed Hudi MetaServer for centralized metadata, visual management, unified permission control, TTL, expression payloads, and multi‑active scaling, and outlines future enhancements such as LLS, multi‑table fusion, and JDBC support.
This article explains Apache Hudi’s write‑side indexing, detailing the indexing API, various index types—including simple, Bloom, bucket, HBase, and record‑level indexes—and their mechanisms, helping readers understand how Hudi validates record existence and optimizes updates and deletions.
This article explains the complete Apache Hudi write pipeline, detailing each step from client creation to commit, and describes the various write operations such as Upsert, Insert, Bulk Insert, Delete, Delete Partition, and Insert‑Overwrite, providing a comprehensive overview for data‑lake practitioners.
This article introduces Tencent's Big Data Processing Suite (TBDS) and the open‑source Gravitino project, explaining how they provide a unified metadata service and a comprehensive, extensible permission model to address data and permission islands across heterogeneous Hadoop and MPP ecosystems.
This article examines the current evolution of data lakes, detailing their overall architecture, batch and real‑time integration methods, Lakehouse core functionalities such as enhanced DML, schema evolution, ACID support, and open‑design principles that enable multi‑cloud deployment and seamless interaction with diverse compute engines.
Open Data Formatsbatch processingbig data architecture
This article provides a practical deep‑dive into StarRocks and Apache Paimon, covering data‑lake fundamentals, the technical advantages of both platforms, performance gains over traditional engines, step‑by‑step migration strategies, deployment options on Alibaba Cloud EMR, and future roadmap plans.
This article presents a practical overview of StarRocks and Apache Paimon data‑lake capabilities, explains their performance advantages, details migration strategies from Trino/Presto and other engines, describes cluster‑to‑cluster migration, and outlines future roadmap for integration and optimization.
The two‑day DataFunCon 2024 Beijing conference gathered hundreds of big‑data and AI experts to discuss the evolution from data lakes to lake‑warehouses, large‑model development, practical applications, and future strategies for enterprises, while showcasing partner exhibitions and a vibrant community spirit.
This article explains how Apache Hudi integrates with Spark to read data, detailing the Spark‑SQL planning stages, the Spark‑Hudi read workflow, and the four main Hudi query types—snapshot, read‑optimized, time‑travel, and incremental—along with example SQL commands and code snippets.
This guide demonstrates how to use Alibaba Cloud's EMR Serverless Spark and Flink Serverless services together with Apache Paimon to ingest streaming data, perform interactive queries, and schedule offline compaction jobs, creating a unified real‑time and batch data lake solution.
This article presents a comprehensive overview of modern Data+AI data lake challenges and solutions, covering the evolution of data lakes, an introduction to Apache Iceberg, practical use of PyIceberg for AI training and inference pipelines, and advanced vector table and indexing techniques for efficient similarity search.
This article introduces Apache Hudi’s storage format, explaining the table layout, metadata and data file organization, the naming conventions of timeline actions, and the trade‑offs between Copy‑on‑Write and Merge‑on‑Read table types for transactional data lakes.
This article presents a comprehensive overview of the evolution from traditional Lambda‑based real‑time data warehouse solutions to a data‑lake‑integrated architecture, detailing the shortcomings of legacy designs, the iterative improvements made at JD Technology, and the technical and operational challenges encountered during implementation.
This article outlines ten critical components of a modern data lake reference architecture for AI/ML, detailing each function, the supporting vendor tools and open‑source libraries, and how they enable scalable storage, MLOps, distributed training, model hubs, vector search, and data visualization.
Apache Iceberg offers a modern, ACID‑compliant table format for data lakes with features like hidden partitions and schema evolution, while StarRocks provides high‑performance query acceleration, metadata caching, and distributed planning to address Iceberg’s latency challenges, enabling seamless lake‑warehouse integration and real‑time analytics.
This article presents a complete guide to Hudi-based real-time data lake ingestion, covering overall data integration architecture, batch and streaming ingestion strategies, advanced table design, and practical recommendations for handling challenges such as deduplication, latency, partitioning, and performance optimization.
This article presents UCloud's USDP‑based streaming data lake warehouse solution that leverages Flink for real‑time processing and Paimon for lake storage, detailing its architecture, advantages, practical scenarios, and providing complete SQL and Flink CDC code snippets for end‑to‑end implementation.
This article introduces Delta Lake 3.1, detailing its release background, the addition of Deletion Vector to Update and Merge commands, metadata‑driven count/min/max optimizations, the Universal Format for cross‑engine compatibility, and a comparative evaluation with Iceberg and Hudi.
iQIYI replaced its costly Lambda architecture with a unified Iceberg‑based lakehouse that combines Flink streaming and batch processing, cutting data latency from hours to minutes, supporting thousands of tables via a multi‑table sink, guaranteeing completeness, and saving millions of RMB in operational costs.
This article explains how to build a high‑performance data‑lake analytics stack by combining StarRocks with Apache Paimon, covering direct queries, Data Cache acceleration, and asynchronous materialized views, and presents benchmark results that show StarRocks achieving up to 4.3× faster query speeds than Trino and significant latency reductions with caching and materialized views.
This article presents a comprehensive overview of JD Tech's real‑time data warehouse evolution, detailing the legacy Lambda‑based design, its shortcomings, the transition to a data‑lake‑integrated architecture, iterative improvements, encountered technical and non‑technical issues, and future outlooks.
OPPO's data architects combined their self‑developed Shuttle service with Alluxio to double performance, halve system pressure, and double throughput, while building a unified Data & AI lakehouse that integrates structured and unstructured data, metadata management, real‑time ingestion, and cloud cost reductions.
iQIYI migrated hundreds of petabytes of Hive tables to Apache Iceberg using dual‑write, in‑place, and CTAS strategies, combined with partition pruning, Bloom filters, and Trino/Alluxio optimizations, achieving up to 40% lower query latency, simplified pipelines, and faster, cost‑effective data lake operations.
Xiaohongshu’s data‑warehouse team integrated Apache Iceberg‑based data‑lake techniques into its existing warehouse, replacing the legacy Hive/Spark stack with global sorting, Z‑order, and upsert‑enabled tables, which cut query latency by up to 90 %, boosted data freshness by 50 %, slashed storage costs by 83 % and saved tens of thousands of GB‑hours of compute daily.
This article introduces a new lakehouse analytics paradigm by combining StarRocks and Paimon, covering the evolution of data lake technologies, key integration scenarios, core technical mechanisms such as JNI connectors, materialized views, and future roadmap for enhanced lakehouse capabilities.
In 2022 Tencent's A/B Test team created the overseas SaaS product ABetterChoice, abstracting internal experiment capabilities, adapting to multi‑cloud compliance, and unifying computation with StarRocks, enabling game titles like Honor of Kings, PUBG Mobile, and Ubisoft to run scalable, compliant A/B experiments worldwide.
This guide walks you through a 5‑minute challenge that shows how to enable Hologres real‑time lakehouse capabilities, configure OSS, DLF, and Hologres services, create external and internal tables, run TPCH Q11 queries, and submit results for prizes.
This article explains what a data lake is, its origins, key characteristics such as collecting all data, enabling diverse user access, and flexible processing, compares it with traditional data warehouses, discusses cost advantages, potential pitfalls like data swamps, and outlines best‑practice considerations for enterprise adoption.
This article provides an in‑depth overview of Apache Paimon as a streaming lakehouse, explains its core features, file layout, consistency guarantees, and offers detailed guidance on integrating and tuning Paimon with Apache Flink for both write and read performance, multi‑writer concurrency, table management, and bucket rescaling.
The 2023 Yunqi Conference in Hangzhou showcased the latest advances in cloud computing and big‑data technologies, examined the evolution from big‑data 1.0 to 3.0, discussed the key difficulties of making big data cloud‑native, and presented a practical case study of MiHoYo’s cloud‑native transformation.
The article describes how NetEase News' recommendation team replaced a slow, batch‑oriented data‑warehouse pipeline with a Flink‑based, Apache Paimon real‑time wide‑table solution that supports partial updates, reduces latency from hours to minutes, and lowers processing costs while handling both deduplication and non‑deduplication recommendation scenarios.
This article explains how Apache Paimon simplifies CDC data lake ingestion with one‑click, low‑cost, low‑latency pipelines, details its architecture and tag‑based Hive compatibility, provides best‑practice configurations, and presents benchmark results showing Paimon outperforming Hive and Hudi in both write and query performance.
Data lakes and data warehouses serve different purposes in big‑data architectures; this article explains their definitions, core attributes, five major distinctions—including data retention, type support, user coverage, adaptability, and insight speed—and offers guidance on selecting or combining the two approaches.
This article introduces Amoro, a lakehouse management platform built on Apache Iceberg, explains why Webex adopted it to overcome Hive limitations, details its AWS GlueCatalog and S3 integration with DynamoDB lock management, and provides step‑by‑step Helm‑based deployment instructions on Kubernetes.
This article details a Chinese tech company's migration of its internal big‑data analytics platform to StarRocks’ compute‑storage separation architecture, describing the original multi‑component setup, the pain points encountered, the evaluation methodology, performance and cost benchmarks, operational optimizations, migration steps, and future roadmap.
This article compares traditional data warehouses, modern data lakes, and emerging lakehouse architectures, explaining their design patterns, advantages, disadvantages, and suitable use cases, while detailing implementation considerations such as schema design, ETL/ELT processes, file formats like Delta, Iceberg, and Hudi, and factors influencing platform selection.
This article presents a comprehensive overview of a unified data lake architecture, evaluates three ingestion solutions, details real‑time ingestion optimizations for Flink‑Hudi pipelines, and describes how Kyuubi enables unified query access across multiple engines, offering practical guidance for large‑scale data processing.
This article explains why Kuaishou built a data lake, outlines the shortcomings of its previous Lambda architecture, describes the adoption of Apache Hudi for unified batch‑stream processing, and details the five major technical challenges and the corresponding solutions implemented to improve performance, consistency, and operational reliability.
ByteDance revamped its traditional Lambda architecture for e‑commerce traffic data by introducing a new lake ingestion solution that reduces development and operational costs, ensures timely and stable data, and outlines future plans covering business background, ODS lake design, archiving tags, delayed data handling, and real‑time stability.
The article provides a deep technical overview of StarRocks 3.0’s data‑lake analysis capabilities, its unified Lakehouse architecture, Catalog integration, Trino compatibility, extensive I/O optimizations, materialized view features, resource isolation techniques, real‑world use cases, and future development directions.
This presentation introduces Iceberg's core capabilities, details Xiaomi's practical applications—including log ingestion, near‑real‑time warehousing, offline challenges, column‑level encryption, and Hive migration—and outlines future development directions such as materialized views and cloud migration, providing a comprehensive view of modern data‑lake engineering.
This article explains why traditional centralized data lakes hinder modern software development, introduces the data‑mesh concept as a decentralized alternative, and walks through an e‑commerce microservice example with concrete steps, data‑API designs, and migration tactics to transition from a monolithic lake to a distributed data mesh.
iQIYI’s data‑middle‑platform team built a four‑zone data lake—raw, product, work, and sensitive—integrated with unified ODS/DWD/MID layers, a metadata catalog, and self‑service tools, leveraging HDFS, Hive/Iceberg, Spark/Trino, and Flink, migrated to Apache Iceberg for real‑time freshness, and now aims to further streamline modules and adopt new technologies.
OrangeFS is Didi’s cloud‑native, multi‑protocol distributed data‑lake storage system that unifies POSIX, S3 and HDFS access on a single logical hierarchy, integrates with Kubernetes via a CSI plugin, supports on‑premise and public‑cloud backends, provides multi‑tenant isolation, and dramatically improves elasticity, utilization and latency for petabyte‑scale workloads such as ride‑hailing logs, machine‑learning training, finance and analytics.
This article explains how StarRocks 3.0 extends real‑time data‑warehouse capabilities to support data‑lake analysis, external catalog integration, Trino compatibility, extensive I/O optimizations, and powerful materialized‑view features that together enable a unified, cloud‑native Lakehouse solution with high performance and flexible resource isolation.
iQIYI now relies on Apache Spark as its main offline engine, processing over 200 000 daily tasks for ETL, data synchronization and analytics, while recent optimizations—dynamic resource allocation, adaptive query execution, compression, rebalance, Z‑order and resource‑governance—have cut compute usage by ~27 %, storage by up to 76 % and improved query speed, completing a large‑scale migration from Hive and paving the way for Spark 3.4 and Iceberg support.
This article details Bilibili's implementation of a streaming data lake using Hudi and Flink, covering background challenges, four case studies, batch‑stream integration optimizations, infrastructure and kernel enhancements, and future work directions.
This article presents a comprehensive technical overview of Alibaba Cloud AnalyticDB MySQL and its Serverless Spark integration, detailing architecture, core optimizations, security enhancements, and real‑world case studies that demonstrate how to achieve cost‑effective, high‑performance data lake analytics.
StarRocks 3.1 introduces a cloud‑native, lakehouse‑oriented architecture with enhanced storage‑compute separation, up to 3‑6× faster data‑lake queries than Trino/Presto, expanded Iceberg and Paimon support, richer materialized view capabilities, new random bucketing, expression partitioning, generated columns, and spill‑to‑disk stability, all backed by extensive performance optimizations and open‑source contributions.
This article presents a comprehensive overview of Impala's high‑performance MPP query engine, its architecture for data‑lake workloads, and detailed performance optimizations including Iceberg table format improvements, manifest caching, and various Codegen techniques such as asynchronous compilation and caching.
To tackle the massive data, high‑throughput and low‑latency demands of large‑model training and inference, the talk proposes a unified data‑lake built on scalable object storage combined with an acceleration layer—either a parallel file system or cloud‑native RapidFS cache—demonstrating multi‑fold training speedups, faster checkpoint uploads, and linear inference scaling.
AI Model StorageAccelerated FilesystemObject Storage
Guandata’s R&D leader outlines how their analytics platform leverages Delta Lake and Spark to deliver fast, ACID‑compliant BI and AI workloads, detailing architecture, key features like schema evolution and time travel, and practical performance tricks such as compaction, vacuuming, and multi‑engine integration.
This article introduces Apache Doris’s data lake federation capabilities, detailing its lake‑warehouse integration design, supported data sources such as Hive, Iceberg, Hudi, and Elasticsearch, performance optimizations for metadata and file access, case studies, community roadmap, and Q&A on replacing Presto.
This article explains how metadata serves as essential infrastructure for data platforms, detailing Huawei's classification framework, governance challenges, management architecture, integrated modeling, data lake handling, service management, and data map construction to bridge business and IT domains.
This article presents a comprehensive overview of Lakehouse‑based in‑lake warehousing, covering common data‑lake misconceptions, the evolution from databases to data warehouses and lakes, the advantages of Lakehouse over traditional architectures, a reference multi‑layer architecture, typical use cases, challenges, future plans, and a brief Q&A.
This article summarizes a technical presentation on how Delta Lake is integrated into a BI+AI platform, covering the product background, data‑lake architecture, Delta Lake features such as ACID transactions, schema management, multi‑engine support, performance optimizations, and future development directions.
This article examines the performance challenges of cloud‑native data lakes built on flat object storage and explains how a hierarchical‑namespace design improves directory operations, reduces request amplification, and delivers significant speedups for big‑data and AI workloads.