The Big Data & AI Product & Technology Monthly Newsletter for April 2026 covers the latest technical updates, product and feature releases, and customer success stories across Alibaba Cloud's Big Data & AI portfolio.
DLC Lingjun Storage Flexible Mount v1.0 is now available. It automatically detects the network characteristics of storage and compute resources, enabling seamless switching between VPC and VCS network mounts. This addresses scenarios where Lingjun compute and CPFS storage reside in different HPN Zones, where cross-zone access is blocked by default.
New GPU resource utilization metrics for DLC tasks are now available. Based on GPU compute usage data (reflecting the percentage of time GPUs spend executing computations in a busy state), these metrics measure task utilization ratios, helping customers manage compute resources more efficiently.
DLC now supports task templates for distributed training in experimental scenarios. Users can pre-fill configuration values, lock critical and error-prone settings, and define task description paradises within templates. This captures task submission experience, organizes tasks within the same experimental direction, and improves task and model iteration efficiency.
EAS now supports dynamic parameter adjustment for services, enabling hot updates to service configurations. Parameters can be added, deleted, modified, and queried in real time without restarting the service or its instances.
DSW now provides a Linux graphical desktop environment based on TurboVNC + noVNC. Users can access a full graphical desktop directly through their browser with no additional configuration required. This meets the needs of development scenarios that require a graphical interface, such as autonomous driving simulation and embodied AI visual debugging.
DSW DockerBoard is now officially available on Lingjun clusters. It supports visual creation and management of multiple secondary sub-containers within a single DSW instance, meeting the needs of enterprise customers (such as Zivara) where multiple developers share a single dev machine. This improves resource utilization and collaboration efficiency.
Xiao PAI now exposes its existing diagnostic and Q&A capabilities as an MCP service, enabling integration into the Agent ecosystem. Additionally, new capabilities include DLC task restart count queries, restart reason summary analysis, and node blacklist status queries — helping users quickly identify the root cause of training task anomalies.
To align with the latest lakehouse architecture developments, MaxCompute will adjust its pay-as-you-go billing model for external tables and data lake analytics computing, effective June 30, 2026 (Beijing Time). This adjustment aims to cover more data sources and reflect true computing costs.
Hologres has added the Japan (Tokyo) 3-AZ Region E with Serverless support across all sales regions, including:
China North 2 (Beijing) / China North 3 (Zhangjiakou) / China North 6 (Ulanqab) / China East 1 (Hangzhou) / China East 2 (Shanghai) / China South 1 (Shenzhen) / China Southwest 1 (Chengdu) / China (Hong Kong) / Japan (Tokyo) / Singapore / Malaysia (Kuala Lumpur) / Indonesia (Jakarta) / South Korea (Seoul) / US (Silicon Valley) / US (Virginia) / Germany (Frankfurt) / UAE (Dubai)
Task priority is now supported in Serverless Spark. Users can assign higher priority to critical business tasks, ensuring they are scheduled first when resources are constrained, thereby maintaining core business stability.
AI Function now supports the qwen3.5-plus, qwen3.6-plus, and tongyi-embedding-vision-plus models, expanding multimodal data processing capabilities. It supports inference, vectorization, and intelligent analysis across multiple input modalities including images, videos, and text.
DuckDB session support is now available, allowing users to perform lightweight data querying and analysis directly within the platform using DuckDB. This is ideal for interactive exploration and rapid analysis of small to medium-sized datasets.
Ray cluster resource mode is now available, enabling users to create and manage Ray clusters for multimodal data processing and distributed Python parallel computing.
StarRocks integration is now supported in the data catalog. Users can add StarRocks to the catalog to access and query related data.
New features include: compatibility with StarRocks Community version 3.5 and above; Stella Lake Optimizer delivering over 200% QPS improvement; high-concurrency stream load server-side Merge Commit for significantly enhanced real-time write throughput; and Cluster Snapshot for cluster-level metadata backup to OSS, meeting enterprise-grade disaster recovery requirements.
EMR Serverless StarRocks now provides business insights with example-driven application scenarios. Insights cover the previous day (T+1) and include query insights, import insights, data table insights, compaction insights, and cache insights.
The EMR AI Assistant is now directly accessible within StarRocks Manager, enabling real-time slow SQL diagnostics, runtime error analysis, and more.
The EMR Serverless Spark Management Skill is now available, enabling users to manage workspaces and submit, query, and manage Spark jobs through natural language interaction, lowering the barrier to entry and improving data development efficiency.
Compaction tasks are now decoupled from business compute groups and run on a dedicated Compaction Service, achieving workload isolation, elastic scaling, and performance optimization. This feature is available on version 3.5.
The latest version 9.3 introduces advanced features including Agent Builder, Workflow, and DiskBBQ, with significant performance improvements across multiple scenarios.
When data is updated, vector fields whose source field values have not changed will directly read cached vector values, avoiding redundant vector computation and reducing resource waste and cost.
A new fused vector module supports selecting multiple fields with multimodal model services to output a single vector value. Joint encoding of text and image fields satisfies both semantic intent and visual style matching requirements, improving retrieval effectiveness.
New features include: Video Summary — leverages large model capabilities to generate summaries, titles, and tags for specified video clips; Video Segmentation — extracts key frames through frame capture and analysis, segmenting videos into corresponding paragraphs.
Milvus now provides built-in Embedding model service capabilities, offering managed Embedding inference services.
Milvus now supports Skills for cluster management and business validation scenarios.
Alibaba Cloud E-MapReduce: Serverless Open-Source Big Data Platform
15 posts | 0 followers
FollowAlibaba Cloud Big Data and AI - April 13, 2026
CloudSecurity - February 11, 2026
CloudSecurity - April 15, 2026
CloudSecurity - March 16, 2026
Alibaba Cloud Big Data and AI - January 21, 2026
Academy Insights - December 25, 2025
15 posts | 0 followers
Follow
Hologres
A real-time data warehouse for serving and analytics which is compatible with PostgreSQL.
Learn More
Big Data Consulting for Data Technology Solution
Alibaba Cloud provides big data consulting services to help enterprises leverage advanced data technology.
Learn More
MaxCompute
Conduct large-scale data warehousing with MaxCompute
Learn More
Big Data Consulting Services for Retail Solution
Alibaba Cloud experts provide retailers with a lightweight and customized big data consulting service to help you assess your big data maturity and plan your big data journey.
Learn MoreMore Posts by Alibaba Cloud Big Data and AI
