[In preview] Public Preview: Azure Container Apps serverless GPUs now support Azure AI Foundry models

-


Overview of Azure Container Apps Serverless GPUs with Azure AI Foundry Models

Azure has announced a public preview for Azure Container Apps serverless GPUs, which now support Azure AI Foundry models. This integration allows users to deploy ready-to-use AI models directly during container app creation, providing flexibility and efficiency in AI application development.

Key Features

  • Azure AI Foundry Models Support: Azure Container Apps now supports deploying Azure AI Foundry models, allowing users to integrate these models during app creation for efficient AI workflows[1][3].
  • Serverless GPUs: These GPUs enable automatic scaling, optimized cold start, and per-second billing. Users only pay for the GPU compute they use, and the feature scales down to zero when not in use[2][4].
  • Benefits for AI Development: Serverless GPUs accelerate AI development by focusing on core AI code rather than infrastructure management. This feature provides a middle layer between Azure AI Model Catalog's serverless APIs and custom models on managed compute[2][4].
  • Data Governance: Data never leaves the container boundary, ensuring full data governance and security while providing a managed serverless platform[2][4].

How It Works

  1. Access to GPUs: Users must request GPU quotas through a customer support case to access serverless GPUs[2][5].
  2. Workload Profiles: Serverless GPUs are supported only for Consumption workload profiles, not for Consumption-only environments[2][5].
  3. NVIDIA GPUs: Users can choose between NVIDIA A100 and T4 GPUs for their applications[2].

Recent Developments

  • General Availability: Azure Container Apps Serverless GPUs are now generally available, providing a robust platform for AI workloads with automatic scaling and per-second billing[4].
  • NVIDIA NIM Integration: Supports NVIDIA NIM microservices for secure and scalable AI model inferencing[4].
  • Dynamic Sessions: Early access is available for Serverless GPU in Dynamic Sessions, enabling running untrusted AI-generated code within compute sandboxes protected by Hyper-V isolation[3].
อ้างอิงค์ : https://azure.microsoft.com/updates?id=492104

ความคิดเห็น

แสดงความคิดเห็น

โพสต์ยอดนิยมจากบล็อกนี้

Retirement: MICROSOFT AZURE OPERATED BY 21VIANET - Action required to migrate to Azure China North 3 region by July 1, 2026 due to China North 1 and China East 1 region retirement

-
รูปภาพ
Retirement of Azure Regions in China Microsoft Azure operated by 21Vianet is planning significant changes to its regional infrastructure in China. Specifically, the China North 1 and China East 1 regions are scheduled for retirement. Key Points About the Retirement Retirement Date : Both China North 1 and China East 1 regions will be retired on July 1, 2026 [1]. Impact : Customers with services in these regions will experience disruptions if they do not migrate their services to other regions by the retirement date[1]. Recommended Action : Microsoft advises customers to migrate their services to alternative regions, such as China North 3 , to avoid service disruptions. Additional Service Updates Beyond the region retirement, there are other service updates relevant to Azure in China: Azure Batch Service Retirement : The Azure Batch service will also be retired on March 31, 2026 , in regions like India West , China North 1 , and China East 1 , due to regional constraints...
อ่านเพิ่มเติม

[Launched] Generally Available: Private subnet

-
รูปภาพ
Here is a structured summary of the Generally Available: Private Subnet launch announcement and key information from the official Azure Updates page referenced (https://azure.microsoft.com/updates?id=492953): General Availability of Azure Private Subnet Announcement Date: May 2025[1] Feature Status: Launched (now generally available for all Azure customers) Description: The "private subnet" feature for Azure Virtual Network is now production-ready, enabling customers to create subnets with enhanced network isolation and security. Key Benefits: Enhanced Security: Private subnets help restrict outbound internet access, reducing the risk of unauthorized data exfiltration. Explicit Internet Access: VMs in a private subnet do not have default outbound internet access, requiring explicit configuration of NAT gateways or other routes for connectivity[5]. Greater Control: Organizations gain more granular management over their Azure VM network traffic, allowing them t...
อ่านเพิ่มเติม

[In preview] Public Preview: Perth - Azure Extended Zones

-
รูปภาพ
Azure Extended Zones in Perth: Public Preview and Updates Overview of Azure Extended Zones Azure Extended Zones are designed to support latency-sensitive and throughput-intensive applications by positioning them closer to end-users, thereby reducing latency and improving overall service delivery[3]. This technology is particularly beneficial for applications requiring quick data processing and fast response times. Public Preview: Perth - Azure Extended Zones The Azure Extended Zone in Perth is currently in public preview, with a planned full deployment by mid-2025[1][4]. This initiative is part of Microsoft's broader strategy to expand its global data center footprint and enhance cloud services in Western Australia[1][5]. Key Features and Benefits Reduced Latency : By hosting applications closer to users, Azure Extended Zones reduce latency, improving the efficiency and reliability of essential services[1][5]. Localized Services : This technology allows data to be processe...
อ่านเพิ่มเติม