Three Cloud Myths I Believed Before Studying Azure Fundamentals
While preparing for AZ-900, I realized that several assumptions I had about cloud computing were incomplete. ☁️ Myth 1: The cloud is always cheaper Cloud services can reduce upfront costs, but unused resources, incorrect sizing, and poor planning can quickly increase the bill. 🖥️ Myth 2: The cloud means there are no servers The physical infrastructure still exists. What changes is how much of it is managed by the cloud provider. ⚙️ Myth 3: More control is always better Additional control also means additional responsibility for patching, monitoring, configuration, security, and maintenance. My biggest takeaway was that cloud computing is not automatically cheaper, simpler, or better. It is about finding the right balance between cost, control, responsibility, and speed. For beginners interested in exploring these concepts, Microsoft Learn provides an official learning path: https://learn.microsoft.com/en-us/training/courses/az-900t00?wt.mc_id=studentamb_615882 Which cloud misconception did you believe when you first started learning?Enabling the Compliance Security Profile (CSP) for HIPAA on Azure Databricks
Microsoft Architect's: Aladdin Alchalabi aalchalabi, Kiran Raja KiranRaja, Peter Lenges PeterLenges, Jessica Reece jareece, Benjamin Coughtry bcoughtry, Anishek Kamal anishekkamal, Tayo Akigbogun takigbogun, Eric Kwashie ekwashie, Peter Lo PeterLo and Rafia Aqil Rafia_Aqil Peer Reviewed: Ted Kim tedkim and Arvind Periyasamy ArvindPeriyasamy Purpose and WHY Azure Databricks has put in place controls to meet the unique compliance needs of highly regulated industries. The requirement for the compliance security profile (CSP) is a joint effort between Microsoft and Databricks for Azure-Databricks workspaces. The value proposition of the compliance security profile is that it provides Customers significantly more hardening and security features. Mandatory Deadline: The Compliance Security Profile (CSP) becomes mandatory for processing HIPAA, HITRUST, and IRAP regulated data on Azure-Databricks by September 1, 2026. Enabling CSP on Workspaces These requirements are checked and enforced on new workspaces today, with enforcement on existing workspaces expected in the future; where prerequisites are missing, clusters may fail to start. Prerequisite Requirement Costs There is a 10% cost of the Azure Databricks product spend within each workspace where CSP is enabled. **Review with your account team for any grace period during which the Enhanced Security & Compliance (ESC) add-on is available at no charge. After the grace period ends, a 10% DBU upcharge applies. Enhanced Security & Compliance add-on For existing workspaces: From Azure portal, click the Settings > Security & compliance on an existing Azure Databricks workspace: **Review Note #2 below Azure VNet encryption Azure Virtual Network encryption must be enabled on the Azure Databricks workspace VNet. Infrastructure as code: Update the encryption block on your VNet resource. In Terraform, that's azurerm_virtual_network. Azure portal: Toggle encryption on the VNet (Overview → Properties → Encryption). Command line: Enable it with the Azure CLI or PowerShell. **Review Note #4 below Supported VM instance types Use a VM series that supports VNet encryption and verify compatibility before enabling the profile. **This does not apply to serverless compute. NOTE: Confirm your workspace is using Premium Pricing tier. The profile can be enabled when a workspace is created or on an existing workspace, through the Azure portal, the Azure CLI, PowerShell, an ARM template, or Terraform. Only the Public Preview, Private Preview, and Beta features listed in this section are supported for workspaces with the compliance security profile enabled: Compliance security profile - Azure Databricks | Microsoft Learn Currently, the compliance security profile checks and enforces only the use of specific VM instance types, not the enablement of Azure Virtual Network encryption. Enforcement of the Azure Virtual Network encryption requirement begins on February 1, 2027, including on workspaces that already have the compliance security profile enabled. This flexibility shall allow customers more time to set up VNET encryption. This has been updated in documentation today (See ‘Important’ box). Regarding rollback, CSP can be reversed via a support ticket, if no regulated data has been processed on a particular workspace. A closer look at VNet encryption CSP is enabled per Databricks workspace, but VNet encryption is applied at the VNet level. Enabling it for a Databricks workload therefore affects every resource within that VNet, not just the workspace. A common approach in a hub-and-spoke design is to leave the hub VNet unencrypted and encrypt only the spoke VNet. The hub typically holds shared services such as the DNS resolver, while the spoke hosts the Databricks workspaces that require CSP. What does it mean for a VNet to be encrypted? An encrypted VNet is a security measure that protects VM-to-VM traffic. Data is encrypted in transit through a DTLS tunnel. This is platform-level encryption, applied automatically to traffic within your VNet and across peered VNets. It requires no changes to your operating system or applications. What happens to my VM-to-VM traffic? Qualifying VM-to-VM traffic is encrypted. Traffic involving unqualified instances simply keeps flowing unencrypted. The only enforcement available today is AllowUnencrypted. Important clarifications Encrypting the VNet does not guarantee all traffic within it will be encrypted. The only traffic that gets encrypted is VM-to-VM traffic where both the source and destination VMs are (1) on a supported SKU and (2) have Accelerated Networking enabled on the network interface. Encrypting the VNet does not drop or break traffic from unsupported SKUs. The only supported GA setting today is to allow unencrypted traffic, so non-qualifying traffic is still permitted; it just isn't encrypted. A future DropUnencrypted setting will drop that traffic instead for further hardening. It isn't available yet, and it's currently unknown whether it will become a required setting for CSP. Review the following recommended steps The steps below represent a validated implementation pattern. The exact network design can vary by environment, but the same prerequisite, isolation, and end-to-end validation principles should be applied. Validated implementation step Recommended approach and expected outcome Isolated sandbox workspace Enable CSP first in a representative non-production workspace. This avoids irreversible changes to DEV or production while the network topology, dependencies, VM compatibility, and operational behavior are validated. Enable CSP and select HIPAA Enable the Compliance Security Profile and select HIPAA under Settings > Security & compliance before processing PHI after September 1, 2026. Enable VNet encryption Enable VNet encryption. **Review Azure Virtual Network encryption limitations: What is Azure Virtual Network encryption? - Azure Virtual Network | Microsoft Learn Start a classic cluster Confirm that a classic cluster starts successfully after CSP and VNet encryption prerequisites are applied. This validates that the selected compute path and VM types remain operational. Validate storage connectivity Confirm storage connectivity continue to work. Confirm rollout readiness Proceed to DEV and production only after the complete private connectivity path, cluster startup, storage access, DNS resolution, data pipelines, and performance have been validated from end to end. Things to Review Enablement is permanent Enabling the compliance security profile, or adding a compliance standard, is intended to be a permanent change. You cannot remove the profile or an individual standard from a workspace that has ever processed regulated data; to revert, you must delete the workspace and create a new one. Validate the configuration in an isolated, representative non-production workspace before enabling DEV or production. Inventory and Assessment Identify Regulated Workspaces: Catalogue all existing Azure-Databricks workspaces. Determine which ones currently process, or are planned to process, data subject to HIPAA, HITRUST, or IRAP. Review Data Pipelines: Map out all data ingress and egress points for these identified workspaces, including connections to on-premises data sources, other cloud services, and external APIs. This helps identify potential network impacts. Verify Prerequisites Before Rollout: Confirm that selected VM instance types support VNet encryption and that every required CSP and networking setting is in place, because missing prerequisites can prevent clusters from starting. Enablement Method: Choose the appropriate tooling for enablement of Azure Portal, Azure CLI, PowerShell, ARM templates, or Terraform to ensure consistency and automation. Keep sensitive data out of customer-defined fields You are solely responsible for ensuring that PHI or other sensitive information is never entered into customer-defined input fields. These include workspace names, compute and resource names, tags, job names, job run names, network names, credential names, storage account names, and Git repository IDs or URLs, all of which may be stored, processed, or accessed outside the compliance boundary. What Changes After Enabling Compliance Security Profile On CSP-enabled workspaces, Partner-powered AI features are disabled by default and some assistive features such as Genie Code are also disabled; a workspace admin can re-enable them if required. In addition, only the specific preview features listed in the compliance security profile documentation are supported. No other Public Preview, Private Preview, or Beta feature may be used to process regulated data. Compliance Security Profile (CSP) enhances the security posture of Azure Databricks by enabling a hardened compute image, enhanced security monitoring, and automatic cluster updates. With automatic cluster updates enabled, classic compute resources are periodically updated and may restart during configured maintenance windows, so production schedules should be planned accordingly. Enhanced security monitoring deploys security monitoring agents on supported compute resources and generates logs that security teams can ingest and analyze. When deploying through ARM templates, CSP, enhancedSecurityMonitoring, and automaticClusterUpdate are configurable security and compliance settings that can be specified as part of the workspace deployment. References Compliance security profile: https://learn.microsoft.com/en-us/azure/databricks/security/privacy/security-profile Configure enhanced security and compliance settings: https://learn.microsoft.com/en-us/azure/databricks/security/privacy/enhanced-security-compliance HIPAA, Azure Databricks, Microsoft Learn: https://learn.microsoft.com/en-us/azure/databricks/security/privacy/hipaa What is Azure Virtual Network encryption: https://learn.microsoft.com/en-us/azure/virtual-network/virtual-network-encryption-overview Create a Virtual Network with encryption: https://learn.microsoft.com/en-us/azure/virtual-network/how-to-create-encryption?tabs Hashicorp azurerm_virtual_network: azurerm_virtual_network | Resources | hashicorp/azurerm | Terraform | Terraform Registry
Azure Incident Retrospective - Please register! Session 2 - Tracking ID: ZJV6-SGG-BW8
Join our upcoming live webcast for a transparent discussion about this recent Azure service incident led by our engineering teams. Network connectivity issues in West US Tracking ID: ZJV6-SGG | Impacted: 23 July 2026 Same content presented in both sessions: pick the one that works best for your timezone! What to expect 📚 Understand What happened, how we responded, and what we learned 💬 Ask Live Q&A with our engineering experts throughout the session 🛠 Learn The fixes we've put in place and guidance for workload resiliency Choose your session Same content presented at both times: pick the one that works best for your timezone: Session 1 17:30 UTC Thursday, 27 Aug 2026 Register now → Session 2 05:30 UTC Friday, 28 Aug 2026 Register now → 10:30 AM US Pacific (PDT) 1:30 PM US Eastern (EDT) 6:30 PM London (BST) 1:30 AM +1 Beijing (CST) 3:30 AM +1 Sydney (AEDT) 5:30 AM +1 Auckland (NZDT) 10:30 PM -1 US Pacific (PDT) 1:30 AM US Eastern (EDT) 6:30 AM London (BST) 1:30 PM Beijing (CST) 3:30 PM Sydney (AEDT) 5:30 PM Auckland (NZDT) Our engineering leaders Jamie Gaudette Vice President Azure Networking Cloud+AI Engineering LinkedIn ↗ ⚠️ Prepare before the livestream Read the Post Incident Review (PIR) ahead of time so you can ask any follow up questions during the live Q&A Helpful resources 🔔 Azure Service Health Alerts Get alerts for relevant incidents by setting up notifications via email, SMS, or webhook 🎥 Past Retrospective Recordings Watch recordings of previous retrospective livestreams 📄 Azure Post Incident Reviews Learn more about PIRs and the retrospective programAzure Incident Retrospective - Please register! Session 1 - Tracking ID: ZJV6-SGG-BW8
Join our upcoming live webcast for a transparent discussion about this recent Azure service incident led by our engineering teams. Network connectivity issues in West US Tracking ID: ZJV6-SGG | Impacted: 23 July 2026 Same content presented in both sessions: pick the one that works best for your timezone! What to expect 📚 Understand What happened, how we responded, and what we learned 💬 Ask Live Q&A with our engineering experts throughout the session 🛠 Learn The fixes we've put in place and guidance for workload resiliency Choose your session Same content presented at both times: pick the one that works best for your timezone: Session 1 17:30 UTC Thursday, 27 Aug 2026 Register now → Session 2 05:30 UTC Friday, 28 Aug 2026 Register now → 10:30 AM US Pacific (PDT) 1:30 PM US Eastern (EDT) 6:30 PM London (BST) 1:30 AM +1 Beijing (CST) 3:30 AM +1 Sydney (AEDT) 5:30 AM +1 Auckland (NZDT) 10:30 PM -1 US Pacific (PDT) 1:30 AM US Eastern (EDT) 6:30 AM London (BST) 1:30 PM Beijing (CST) 3:30 PM Sydney (AEDT) 5:30 PM Auckland (NZDT) Our engineering leaders Jamie Gaudette Vice President Azure Networking Cloud+AI Engineering LinkedIn ↗ ⚠️ Prepare before the livestream Read the Post Incident Review (PIR) ahead of time so you can ask any follow up questions during the live Q&A Helpful resources 🔔 Azure Service Health Alerts Get alerts for relevant incidents by setting up notifications via email, SMS, or webhook 🎥 Past Retrospective Recordings Watch recordings of previous retrospective livestreams 📄 Azure Post Incident Reviews Learn more about PIRs and the retrospective programSimplify AKS observability with Azure Native New Relic Service
An Azure Native path to New Relic Intelligent Observability for AKS AKS environments are dynamic by design. Applications can span clusters, namespaces, nodes, pods, and containers, while workloads scale and change continuously. Obtaining consistent visibility often requires platform teams to deploy and maintain monitoring components separately on every cluster. Azure Native New Relic Service simplifies this process by integrating New Relic onboarding and management into Azure. Customers can already use the service to: Create a new New Relic account or link an existing account from Azure. Configure the forwarding of Azure platform metrics and logs to New Relic. View the monitoring status of Azure resources. Consolidate procurement and eligible New Relic charges through Azure Marketplace. Monitor multiple Azure subscriptions through a single New Relic resource. With AKS extension support, customers can now extend this native management experience to their Kubernetes clusters. Install the New Relic integration from the Azure portal The new experience follows the same simple model used by Azure Native integrations for other compute resources. From an Azure Native New Relic Service resource, customers can navigate to New Relic account config > Azure Kubernetes Services, select an eligible AKS cluster, and choose Install Extension. Azure then deploys the New Relic Kubernetes integration by using the AKS cluster extensions framework. After deployment completes, the portal displays the installation status for the cluster. Customers can return to the same experience to review the status or select Uninstall Extension when monitoring is no longer required. AKS cluster extensions provide an Azure Resource Manager-based approach for installing and managing services on AKS. This gives customers a consistent Azure control plane experience for deployment and lifecycle operations instead of requiring a separate, manual Helm installation for each cluster. Gain deeper visibility into Kubernetes workloads The extension deploys the New Relic Kubernetes integration to the selected AKS cluster. The integration provides visibility across Kubernetes infrastructure and workloads, including cluster, node, namespace, deployment, pod, and container health and performance. Depending on the enabled New Relic configuration, customers can also bring together Kubernetes events, logs, and Prometheus-formatted metrics with application and Azure platform telemetry in New Relic. This helps application, platform, and site reliability engineering teams investigate issues across the stack without stitching together disconnected views. Teams can use New Relic to: Understand resource consumption and health across clusters, nodes, pods, and containers. Identify unhealthy workloads, container restarts, and capacity constraints. Correlate Kubernetes infrastructure signals with application performance data. Explore Kubernetes entities and relationships through New Relic's cluster experience. Create dashboards, alerts, and operational workflows using telemetry from Azure and AKS. The result is a more direct path from detecting an issue to understanding its impact on applications and users. Reduce operational toil with unified telemetry For organizations operating multiple AKS clusters, consistency is as important as visibility. Manual installation can lead to configuration drift, missed clusters, and additional work whenever monitoring components need to be changed. The Azure Native New Relic Service experience helps address these challenges by providing: Simplified onboarding: Install the integration from the Azure portal without building a separate deployment workflow. Centralized visibility: Review AKS extension status alongside other Azure resources connected to New Relic. Azure governance alignment: Use Azure Resource Manager and Azure role-based access control as part of the management experience. Lifecycle management: Install or remove the extension through a consistent Azure workflow. Unified, full-stack observability: Connect AKS telemetry with application, infrastructure, log, and Azure platform data in New Relic. This experience is particularly useful for platform teams that want to make observability available as a standardized service while allowing development teams to use New Relic for troubleshooting and performance optimization. Get started To begin monitoring AKS with Azure Native New Relic Service: Either browse to the Marketplace offer listing or in the Azure portal, create an Azure Native New Relic Service resource. Go to New Relic account config > Azure Kubernetes Services. Select the AKS cluster that you want to monitor. Select Install Extension and then confirm the installation. After the status changes to Installed, open New Relic to explore your Kubernetes data and configure the dashboards and alerts appropriate for your environment. If you face any technical challenges do raise a support ticket and share your feedback.
Thursday Architecture Lesson #1
Azure Compute: VM, App Service, Container Apps or AKS? Don't memorize the services. Remember this: V → VM Need maximum control or have a legacy workload. A → App Service Want managed hosting for web apps or APIs. C → Container Apps Want containers without taking on Kubernetes complexity. K → AKS You genuinely need Kubernetes capabilities. 🧠 The Architect's Rule: “Don't choose AKS because you can. Choose AKS because you need it.” The right architecture isn't the one with the most powerful technology. It's the one that gives the business the right balance of: Capability • Cost • Security • Scalability • Operational complexity For example, if a simple web application can run effectively on App Service, introducing Kubernetes may add complexity without adding meaningful business value. The best architects don't ask: “What is the most advanced technology we can use?” They ask: “What is the simplest architecture that meets the requirements?” 💬 Your turn: If you were modernizing a traditional 3-tier application today, which would you choose first — VM, App Service, Container Apps or AKS? And most importantly, why?Find out a server with Azure AD Connect
Hi All, Normally, if someone need to find on which server Azure AD Connect is installed it can be done in Azure portal under Microsoft Entra Connect - Microsoft Entra Connect Health - Sync Services. Is there any way to find out on which server Azure AD Connect is installed if Microsoft Entra Connect Health is not installed or its service stopped? I know in this case no server will be displayed under Microsoft Entra Connect Servers. The reason why I'm asking, someone deployed Azure AD Connect and ran syncing but for unknown reason stopped Azure AD Connect Health service. Because of that I couldn't find on which server Azure AD Connect tool was installed until that person advised. If he didn't tell me I most likely would need to sign in to each production server to check. However, just wondering if it can be done with some PowerShell command. Thanks.AI Gateway tier of API Management now in public preview
Today, we are introducing the AI Gateway tier of Azure API Management, now in public preview. It gives platform teams a purpose-built experience built specifically for AI workloads - publishing and governing models and MCP servers. Controls are configured through policy cards rather than XML and expressions, and the portal experience and control plane are structured around models, MCP servers, and tools rather than APIs. (For brevity, we refer to the AI Gateway tier as AI Gateway throughout the rest of this article.) AI Gateway is built on Azure API Management, bringing proven operational capabilities to AI workloads. The resource runs in your subscription, uses your Entra tenant, and sends telemetry to destinations you control. The operating model will be familiar to existing API Management customers, but the interface is built around AI workloads. The AI Gateway tier is intended for teams that want this focused experience; other API Management tiers remain the right choice when organizations also need general-purpose API management or capabilities not included in the AI Gateway experience. A practical model for platform teams The AI Gateway gives platform teams a shared place to manage models, MCP servers, policies, and observability destinations, with access controlled through Azure RBAC. For example, a central platform group can connect a set of approved models and tools and publish them for application teams. The application teams can test those assets in the test console and build against them without routing every change through the central group. The platform group still owns the shared guardrails and can see how the assets are being used. After an asset is published, developers can create a named runtime key and begin calling the gateway immediately. Bring the models and tools you already use Most organizations don't standardize on a single model provider. Different models are selected based on quality, latency, cost, geography, or specialized capabilities. The preview supports models from Microsoft Foundry including OpenAI, Anthropic, Mistral, and other Foundry hosted models, as well as models hosted in AWS Bedrock, Google Vertex AI, OpenAI, and Anthropic. A guided wizard simplifies importing models from Microsoft Foundry. Other providers can be added by configuring a connection, with backend authentication configured as part of that connection. All published models are available under the same stable endpoint. Applications continue to use supported API formats such as OpenAI Chat Completions and Responses or Anthropic Messages directly or via SDKs. The AI Gateway extends governance beyond models to the MCP servers and tools agents use to interact with enterprise systems. You can expose an existing MCP server over SSE or Streamable HTTP, turn all or selected operations from a REST API into an MCP server by uploading its OpenAPI specification, or use more than 1,400 connector-backed tools from the Power Platform and Logic Apps library. You can also federate multiple MCP servers behind a single server, so an agent connects once and sees the tools across those servers. Backend authentication supports an API key, OAuth client credentials, managed identity, or mTLS. Governance that's built in Organizations need consistent governance across models and MCP servers without requiring every application team to implement those capabilities independently. The AI Gateway portal presents governance policies through an intuitive card-based experience rather than requiring policy XML. The same policies are expressed as JSON properties, making them easy to manage as infrastructure as code and to audit and enforce across a fleet with Azure Policy. In the public preview, those cards cover request and token rate limits, token quotas, Azure AI Content Safety, and fallback to a secondary model. Policies are applied per asset, making it clear which controls protect each model or MCP server. OpenTelemetry-based token metrics The AI Gateway emits token-usage metrics through OpenTelemetry, with attributes following GenAI and cloud semantic conventions. Metrics can be sent to Application Insights, Datadog, Splunk, Grafana Cloud, or another OTLP endpoint. The portal provides a monitoring view over Application Insights data. Better together: Microsoft Foundry and AI Gateway With AI Gateway, teams can extend the same governance controls, for example token rate limits and quotas, across models hosted in Microsoft Foundry and models hosted elsewhere. Foundry and non-Foundry models are published through gateway-managed endpoints, giving applications and agents a consistent way to access governed models regardless of where they are hosted. Foundry-hosted agents can consume curated sets of tools from Foundry toolboxes, with access to the underlying MCP servers and APIs governed through AI Gateway. Together, Microsoft Foundry and AI Gateway cover the enterprise application lifecycle: Foundry for building and running AI applications, and AI Gateway for publishing, governing, and observing models, tools, and MCP servers across your AI estate. The new AI Gateway tier will soon be available through the gateway experience in Microsoft Foundry portal. We are working toward a seamless, integrated AI Gateway experience within Foundry portal and will share more about that work separately. Available today in public preview The AI Gateway tier is available today at no cost in public preview in East US 2 and Sweden Central. Pricing will be shared separately. To provision a resource, add a model or MCP server, and make a first call click this to go to the AI Gateway tier portal and try it. If you prefer to start from code, use a sample to deploy all the required resources for a Foundry-hosted agent configured to access its model and tools through AI Gateway. We look forward to your feedback as we continue to rapidly evolve AI Gateway.