Back to Archive

Azure News - 2026-06-05

2026-06-05
最終更新: 2026-08-27 21:13:49 JST

AI による概要

14 記事

この日は Microsoft Foundry IQ の一般提供が中心でした。プロジェクトごとに検索パイプラインを作り直すことなく、エージェントをエンタープライズデータに接地させるマネージドなナレッジ層を提供します。あわせて Fabric のオントロジーを連携ナレッジソースとして扱う機能、Azure SQL Database を第一級のナレッジソースとする機能がいずれもプレビューで追加されました。Azure AI Search と Foundry ナレッジベースの間のプライベートなエンドツーエンド接続も GA しています。監視では Azure Monitor に Service Level Indicators / Objectives が加わり、OpenTelemetry Collector 経由での OTLP シグナル取り込みも一般提供となりました。AKS では Azure CNI Overlay の Pod CIDR 拡張が GA しています。

主要トピック

  • ナレッジ層: Microsoft Foundry IQ が GA、検索パイプラインを作り直さずにエージェントをエンタープライズデータへ接地

  • ナレッジソース: Fabric オントロジーと Azure SQL Database をナレッジソースとして扱う機能がプレビュー

  • ネットワーク: Azure AI Search と Foundry ナレッジベース間のプライベートなエンドツーエンド接続が GA

  • 監視: Azure Monitor に SLI / SLO が追加、OpenTelemetry Collector 経由の OTLP 取り込みが GA

  • AKS: Azure CNI Overlay の Pod CIDR 拡張が一般提供、IP アドレス設計の自由度が向上

  • 研究開発: Microsoft Discovery が GA、科学・工学分野のエージェンティック AI ワークフローを構築・統制

AI (Claude Opus 5) が生成 · 2026-08-28 09:36:58 JST

Azure Updates

[Launched] Generally Available: Azure Monitor Service Level Indicators (SLI)

Azure Monitor now includes Service Level Indicators (SLIs) and Service Level Objectives (SLOs), giving teams a clearer way to measure how customers actually experience their applications. Instead of relying only on infrastructure signals like CPU or isola

[In preview] Public Preview: Content Understanding chunking and image verbalization in Azure AI Search

Azure AI Search now supports Content Understanding chunking and image verbalization in public preview, expanding the built-in skill that parses rich documents during indexing. Indexers can split documents into semantically meaningful chunks and generate t

[Launched] Generally Available: Private Connectivity for Azure AI Search and Foundry Knowledge Bases

Azure AI Search and Foundry Knowledge Bases now support private, end-to-end network connectivity between search resources and Foundry services. Customers can route ingestion, enrichment, retrieval, and agent traffic over Shared Private Link or Network Sec

[In preview] Public Preview: Fabric IQ Ontology Knowledge Source in Microsoft Foundry IQ

Microsoft Foundry IQ now treats a Microsoft Fabric Ontology as a federated knowledge source in public preview, giving agents a way to query the semantic layer customers already curate in Fabric. Foundry IQ translates natural language questions into ontolo

[Launched] Generally Available: Microsoft Foundry IQ

Microsoft Foundry IQ reaches general availability, giving developers a managed knowledge layer that grounds agents in enterprise data without rebuilding retrieval pipelines for each project. Teams connect SharePoint, OneLake, Azure Blob, and other sources

[In preview] Public Preview: Tool search in Microsoft Foundry toolboxes

Microsoft Foundry adds tool search to Foundry toolboxes in public preview, giving developers and admins a fast way to find the right tool inside large, multi-team catalogs. As your toolbox grows, sending every tool definition on every turn gets expensive.

[In preview] Public Preview: APIM Support for Foundry Models in Azure AI Search

Microsoft Foundry brings Azure API Management support to all Foundry model integrations in Azure AI Search, now in public preview. Enterprise platform engineers and AI solution teams running large-scale RAG pipelines can place Foundry and Azure OpenAI mod

[In preview] Public Preview: Azure SQL as a knowledge source in Foundry IQ

Microsoft Foundry adds Azure SQL Database as a first-class Knowledge Source in Azure AI Search, available in public preview. Enterprise developers and solution architects building Copilot, RAG, and agentic experiences can now expose authoritative tables a

[Launched] Generally Available: Microsoft Discovery

Microsoft Discovery is now generally available, providing research and development organizations an enterprise platform for building and governing agentic AI workflows across scientific and engineering disciplines. Also available in preview is the Microso

[In preview] Public Preview: AI pipelines in Azure HorizonDB

AI pipelines for Azure HorizonDB is now available in public preview. AI pipelines let you describe an AI data ingestion workflow, chunking, embedding, extraction, generation, and ranking, declaratively in SQL, and run it as a fault-tolerant pipeline insid

[Launched] Generally Available: Ingest OTLP signals into Azure Monitor with the OpenTelemetry Collector

Azure Monitor's support for native ingestion of OpenTelemetry Protocol (OTLP) signals is now generally available. This enables you to send telemetry data directly from OpenTelemetry-instrumented applications and platforms to Azure Monitor. You can configu

[Launched] Generally Available: Azure Files assessments are now available worldwide using Azure Migrate

We’re pleased to announce that Azure Migrate now generally supports discovery and assessment of SMB and NFS file shares hosted on Windows and Linux servers. This gives customers a complete, data-driven view of their file share environment and a clear path

[In preview] Public Preview: Voice sync for avatar support in Voice Live API in Microsoft Foundry

In public preview, Microsoft Foundry now supports avatar voice sync with custom voices in the Voice Live API, allowing developers to pair branded or persona-specific text-to-speech voices with real-time avatar experiences. This update helps teams building

Azure Networking Blog

Pod CIDR Expansion Generally Available and IP Address Planning on Azure CNI Overlay

詳細を表示

In networking with Azure CNI Overlay, the cluster-wide pod CIDR is logically partitioned into smaller “node” blocks where each node is assigned a fixed CIDR slice (/24) by Azure. This decouples pod networking from the VNet address space entirely because pods receive addresses from a private CIDR that is separate from the VNet.

By default, Azure CNI Overlay uses a pod CIDR of 10.244.0.0/16 which provides 65,536 addresses. Since each node consumes 256 addresses from the /24 slice, the default cluster has a node scaling limit of 65,536 divided by 256, or 256 nodes. Choosing a pod CIDR at cluster creation effectively sets an upper bound on how many nodes the cluster can accommodate.

Pod CIDR

Per-Node Block

Max Nodes Supported

/16

/24

256

/15

/24

512

/14

/24

1,024

What does Pod CIDR Expansion enable?

Even with careful upfront planning, long-lived clusters grow in ways that are difficult to anticipate. For organizations using Azure CNI Overlay, this previous represents a difficult migration without meticulous IP planning.

Pod CIDR expansion allows you to expand the existing CIDR without downtime or node reimaging. Instead of being locked to the range chosen at cluster creation, operators can expand the available pod address space with minimal operational burden.

Choosing a Pod CIDR

In addition to node scaling limits, there are other considerations for pod CIDR planning:

  • Overlapping pod CIDRs across clusters – even though pod IPs are not directly routable between clusters, overlapping CIDRs can cause problems with observability tooling or cross-cluster networking on top of the overlay. Careful planning can prevent having to recreate the cluster down the road.
  • Accounting for system node pools – each system node also consumes a /24 block. IP address planning should factor nodes running cluster control plane components in addition to existing workloads.

Learn More

Read more about Azure CNI Overlay:

Try pod CIDR expansion:

前日
Archive
翌日