Change the Elastic Pool Storage Size using Azure Monitor and Azure Automation

by Scott Muniz | Sep 2, 2020 | Azure, Microsoft, Technology, Uncategorized

This article is contributed. See the original author and article here.

Introduction to Elastic Pool:
https://docs.microsoft.com/en-us/azure/sql-database/sql-database-elastic-pool

In this article, we will setup an Azure Monitor Alert to Scale up the Storage limit of an SQL Elastic Pool on Azure. Please read more about Elastic Pool in the above article.

We will divide this into three parts

i.   Setting up an Automation Runbook and Webhook
ii.  Setting up an Alert Action Group
iii. Setting up an Alert under Azure Monitor.

We will not talk much about Azure Automation or Azure Monitor as they are off the topic, we will only cover the steps for setting up of this Auto scale of storage. Here are some of the articles that should bring you up to Speed.

Create Azure Automation Account [ In this case, we would need to use RunAsAccount]

https://docs.microsoft.com/en-us/azure/automation/automation-create-standalone-account

Using Az modules in Azure Automation Account

https://docs.microsoft.com/en-us/azure/automation/az-modules

Azure Monitor

https://docs.microsoft.com/en-us/azure/azure-monitor/

Azure Monitor Overview

https://docs.microsoft.com/en-us/azure/azure-monitor/overview

Monitoring Azure Service

https://docs.microsoft.com/en-us/azure/azure-monitor/insights/monitor-azure-resource

Setting Up Automation Modules

By default, you cannot run both Az and Rm Modules on the Automation and this is explained here:

https://docs.microsoft.com/en-us/azure/automation/az-modules

So, we will import the Az Modules to the Automation Account and not Rm Modules. By default, when you create an Automation Account, there is a bunch of modules imported and we will not touch them as we will use the AZ modules.

Here is what you need to do.

1. Go to Azure Automation Account.
2. Click on Modules under Shared resource.

    

3. Click on Browse gallery.
   
   
4. Search for Az.Accounts and Click on Import.

   

5. Likewise, search of Az.sql and import it too [ once the Az.Accounts import is complete. Otherwise, it may fail].
6. Let the modules get imported.
7. Once the modules are imported, you would see the status as available..
   
   
8. Further, you don’t need to add any modules as we will use only SQL related Cmdlets unless you are using this Automation Account for other purposes.
9. Next, we will need to setup an Automation Account Runbook, for that Navigate to Runbooks under Process Automation under Automation Account.

   

10. Click on Create Runbook and Provide the details as below, Click Ok

    

11. In the Runbook Edit section, copy paste the following script:

    #Author: Shashanka Haritsa
    #Date: 19th March 2020
    <#WARNING: The sample scripts are not supported under any Microsoft standard support program or service. The sample scripts are provided AS IS without warranty of any kind.
    Microsoft further disclaims all implied warranties including, without limitation, any implied warranties of merchantability or of fitness for a particular purpose. The entire risk arising out of the use or performance of the sample scripts and documentation remains with you. In no
    event shall Microsoft, its authors, or anyone else involved in the creation, production, or delivery of the scripts be liable for any damages whatsoever (including, without limitation, damages for loss of business profits, business interruption, loss of business information, or other pecuniary loss) arising out of the use of or inability to use the sample scripts or documentation, even if Microsoft has been advised of the possibility of such damages. #>

    #Read Webhook data
    param
    (
    [Parameter (Mandatory = $false)]
    [object] $WebhookData
    )
    # If runbook was called from Webhook, WebhookData will not be null.
    if ($WebhookData) {
    #Authenticate to Azure First using an RunAsAccount
    $getrunasaccount= Get-AutomationConnection -Name ‘AzureRunAsConnection’
    Add-AzAccount -ServicePrincipal -ApplicationId $getrunasaccount.ApplicationId -CertificateThumbprint
    $getrunasaccount.CertificateThumbprint -Tenant $getrunasaccount.TenantId
    # Authentication Complete
    $WebhookData.Requestbody
    $Converteddata = $WebhookData.Requestbody | ConvertFrom-Json
    $resourcegroupname = $converteddata.data.context.resourceGroupName
    $resourceName = $converteddata.data.context.resourceName
    $getservername=(($converteddata.data.context.resourceId) -split(‘/’))[8]
    #Read ElasticPools Current storage and double it
    $GetElasticPoolStorage=(Get-AzSqlElasticPool -ElasticPoolName $resourceName -ResourceGroupName
    $resourcegroupname -ServerName $getservername).StorageMB
    $GetElasticPoolStorage
    $NewStorage = ($GetElasticPoolStorage *2) #I am just Increasing my storage to 100% more for my Standard
    Plan so I am multiplying the storage by 2, you may need to change this according to your requirement
    #Set the new storare limit
    Set-AzSqlElasticPool -ElasticPoolName $resourceName -ResourceGroupName $resourcegroupname –
    StorageMB $NewStorage -ServerName $getservername
    }
    Else{
    Write-output “No Webhookdata found. Exiting”
    }
    

12. Click on save and Click on Publish.
13. Now, we will need to create a Webhook. Under the same Runbook, Click on Webhooks

    

14. Click on Add Webhook.

    

15. Under Create Webhook, give it a name and copy the URL to a safe place from where you can retrieve it in future. [ NOTE: This URL cannot be retrieved after creation, so please keep it safe] Click ok and Click on Create.

    

16. Once the Webhook is created, you will see that under the Webhooks section.

This Completes the first part where we have created the Automation Runbook, setup modules and a Webhook.

Setting up an Alert Action Group

In this section, we will create an Action Group that we will use with an Alert.

Please follow the steps below to create an Action Group

1. Login into Azure Portal [ If you haven’t already]
2. Navigate to Azure Monitor →Alerts and Click on Manage actions

   

3. Next, click on Add action group and fill in the information as needed.
4. Under the Action Name, provide a name as desired and under Action Type, select Webhook

   

5. A Webhook URI screen pops up on the right-hand side, please use the Webhook URL we had copied during the Webhook creation under the Automation Account and click ok.

   

6. Click OK again on the Add action group screen. This will create an action group.

This completes the creation of Action Group.

Setting up an Alert under Azure Monitor

In this part, we will create an Alert that will trigger our Runbook whenever the used space is greater than some value. Please follow the steps below.

1. Navigate to Azure Monitor
2. Click on Alerts and Click on New alert rule
3. Under the resource, click on Select

   

    

4. Filter the Subscription and Resource type as SQL elastic pools and location, select the Elastic Pool of Interest. This should populate the resource as below.

   

    

5. Now, click on add under condition. Select Signal type as Metrics and Monitor Service as Platform
6. Select the Signal name of interest, in this case we will select Data space used percent

   

    

7. Once you select the Metric, you will now need to add alert logic, lets say that you would like to trigger an alert when the Percentage used space is 70 [ Average] for last 1 hour, we will set it up as below:

   

    

   What does it mean? We are checking the Average Data space used Percentage for last one hour and we will evaluate this condition every 5 minutes as a part of Alert.

8. Click on done and now click on Add under ACTIONS GROUPS and select the one you created during the action group creation.
9. Now provide Alert details and a Description. Select Severity of Interest. Once you are happy with the details provided, click Create alert rule

That covers all the three configurations involved. Whenever the data space used percentage on the Elastic Pool increases over 70%, an alert will be triggered, and the Runbook invoked through Webhook will resize the storage on the Elastic Pool.

IMPORTANT NOTE:

• This above sample document is for reference purpose only and is provided AS IS without warranty of any kind.
• The author is not responsible for any damage or impact on the production, the entire risk arising out of the use or performance of the above sample document remains with you
• For the Script section, under Automation Runbook setup, we have taken Standard plan[ Elastic Pool] in account and have only doubled the storage based on our requirement, if your requirement is different, you should evaluate the logic for increasing the storage and then amend the script as necessary.

Accelerating AI for COVID-19 on Microsoft Azure Machine Learning using Clara Imaging from NVIDIA NGC

by Scott Muniz | Sep 2, 2020 | Azure, Microsoft, Technology, Uncategorized

This article is contributed. See the original author and article here.

This post was co-authored by Krishna Anumalasetty, Tom Drabas, Nalini Chandhi from Microsoft and Abhilash Somasamudramath, Manuel Reyes Gomez, Brad Genereaux, Akhil Docca from NVIDIA

Figure 1. CT-Scan Images Classified for COVID-19 Probability

COVID-19 has fundamentally transformed the world we live in. As the scientific community across the globe unites in the face of this pandemic, it is crucial to enable researchers to collaborate and leverage tools that will speed up detection and drug discovery for COVID-19. The power of AI in radiological medical imaging is helping with faster detection, segmentation, and notifications.

Leveraging AI for healthcare applications is challenging for many reasons. The complicated science required to build and train deep learning neural networks, setup and maintenance of the supporting infrastructure required to develop, deploy, and manage these applications at scale, pose barriers towards addressing our most pressing healthcare challenges with AI.

Cloud computing has enabled researchers with easy access to scalable, on-demand infrastructure to get their AI applications up and running quickly. In the medical imaging space, the real need is for a platform that combines the power of NVIDIA GPUs with a secure environment that also allows easy access to AI software. NVIDIA Clara™ is a full-stack GPU-accelerated healthcare framework accelerating the use of AI for medical research and is available on the NVIDIA NGC Catalog.

Azure Machine Learning

Azure Machine Learning (Azure ML) empowers developers, data scientists, machine learning engineers, and AI engineers to build, train, deploy, and manage machine learning models. It is an open platform with built-in support for open-source tools and frameworks, such PyTorch, SciKit Learn and, TensorFlow along with numerous Integrated Developer Environments (IDEs), supporting key  languages like Python and R.

Azure ML abstracts the set-up, installation and configuration of the machine learning environment, saving you the hassle of infrastructure management by  taking care of the underlying technicalities. This enables domain experts, such as healthcare researchers and developers, to build mission-critical AI solutions faster and easier. Whether the project requires image classification, object detection, speech analysis, or natural language processing. Azure ML streamlines AI-powered solution development.

Productivity is also boosted by accelerating your training jobs with Azure’s global infrastructure. Enhance your workflow by scaling out to multi-node compute clusters or scaling up to powerful GPU-enabled machines. Combining this with end-to-end AI lifecycle management through industry-leading MLOps means data science teams can collaborate better and get to production quicker. And, with more than 60 compliance certifications including FedRAMP High, HIPAA and DISA IL5, plus configurable security features, Azure ML allows you to create a trusted working environment.

NVIDIA NGC Catalog and Clara

A key component of the NVIDIA AI ecosystem is the NGC Catalog. It is a software hub of GPU-optimized AI, HPC, and data analytics software built to simplify and accelerate end-to-end workflows. With over 150 enterprise-grade containers, 100+ models, and industry-specific SDKs that can be deployed on-premise, cloud or at the edge, the NGC Catalog enables data scientists and developers to build best-in-class solutions, gather insights, and deliver business value faster. Every single asset in the NGC Catalog is validated for performance, quality, and security by NVIDIA, providing you with the confidence needed to deploy within the Azure ML environment.

The deep learning containers in the NGC Catalog are updated and fine-tuned monthly to enable maximum performance through software-driven optimizations that augment GPU hardware acceleration. These performance improvements are made to libraries and runtimes to extract maximum performance from NVIDIA GPUs. Pre-trained models from the NGC Catalog help speed up the application building process. You can find more than 100 pre-trained models across a wide array of applications such as image analysis, natural language processing, speech processing and recommendation systems. The models are curated and tuned to perform optimally on NVIDIA GPUs for maximum performance. By applying transfer learning, you can create custom models by retraining it against your own data.

NVIDIA Clara is one many artifacts available in the NGC Catalog. Clara is a healthcare framework that comprises of full-stack GPU-accelerated libraries, SDKs, and reference applications to create secure, and scalable healthcare applications.

The Clara family of application frameworks include:

• Clara Imaging: Application frameworks for accelerating the development and deployment of AI based medical imaging workflows in radiology and pathology, as well as in some medical instruments
• Clara Parabricks: Computational framework supporting genomic analysis in DNA and RNA
• Clara Guardian: Application framework and partner ecosystem that simplifies the development and deployment of smart sensors within the hospital with multimodal AI

Clara Imaging, the focus of this blog, offers easy-to-use, domain-optimized tools to create high-quality, labeled datasets, collaborative techniques to train robust AI models, and end-to-end software for scalable and modular AI deployments. It consists of two essential elements – Clara Train and Clara Deploy:

• Clara Train is a framework that includes two main libraries; AI-Assisted Annotation (AIAA), which enables medical viewers to rapidly create annotated datasets suitable for training, and a Training Framework, a TensorFlow based framework to kick start AI development with techniques like transfer learning, federated learning, and AutoML.
• Clara Deploy provides a container-based development and deployment framework for multi-AI, multi-domain workflows in smart hospitals for imaging, genomics, and signal processing workloads. It leverages Kubernetes to enable developers and data scientists to define a multi-staged container-based pipeline.   

Figure 2. The entire Clara pipeline that includes Clara train and deploy

The entire Clara Framework can be easily accessed from the Clara portfolio page. In addition to Clara Train and Deploy, reference models and pipelines are also available for download. Applying transfer learning, developers can create new models with their own custom dataset.

Figure 3. Various types of pre-trained models that cover both 2D and 3D segmentation and classification for different use cases

NGC-AzureML Quick Launch Toolkit

A ready-to-use Jupyter notebook was created to showcase the fine-tuning of a pre-trained COVID-19 CT Scan Classification model from the NGC Catalog.

To help automate the deployment, we have developed the NGC-AzureML Quick Launch toolkit that leverages the Azure ML SDK and creates the necessary compute and software resources needed to run Machine Learning applications. The Azure ML SDK uses Azure Machine Learning Compute Clusters, which require their own quota, and can be created using the same mechanism as the process followed to setup a quota for Azure VMs.

The toolkit can be used to launch any asset from the NGC Catalog, by just changing a config file. In our example, the toolkit takes relevant assets for Clara, but you can customize it to pull assets related to other uses cases such as computer vision, natural language understanding, and more.

The toolkit automates the steps outlined below:

• An Azure instance with NVIDIA GPUs is configured with the right NVIDIA libraries and drivers
• Setting the desired NGC Catalog image (Clara Train SDK in this example) onto the Azure instance
• Uploading additional material: model(s) (the pre-trained COVID-19 CT Scan Classifier in this case), auxiliary code and the corresponding datasets from the NGC Catalog to the Azure instance
• Loading the ready-to-use Jupyter notebook from the NGC Catalog that contains the application
• Installing JupyterLab the Azure instance and accessible locally to run the ready-to-use Jupyter notebook

To set up the AzureML environment, you only need to run two commands in the command line interface (CLI):

Install azureml-ngc-tools

First, install the NGC-AzureML Quick Launch Toolkit on the local machine, via Pip:

pip install azureml-ngc-tools

Configure azure_config.json

This file contains the Azure credentials and desired choice of instance type. A ready-to-use template for this example to complete can be downloaded here, which should be edited with user credentials.

An example of how the azure_config.json file might look is the following:

{
    "azureml_user":
    {
        "subscription_id": "ab221ca4-f098-XXXXXXXXX-5073b3851e68",
        "resource_group": "TutorialTestA",
        "workspace_name": "TutorialTestA1",
        "telemetry_opt_out": true
    },
    "aml_compute"
    {
        "ct_name":"clara-ct",
        "exp_name":"clara-exp",
        "vm_name":"Standard_NC12s_v3",
        "admin_name": "clara",
        "min_nodes":0,
        "max_nodes":1,
        "vm_priority": "dedicated",
        "idle_seconds_before_scaledown":300,
        "python_interpreter":"/usr/bin/python",
        "conda_packages":["matplotlib","jupyterlab"],
        "environment_name":"clara_env",
        "docker_enabled":true,
        "user_managed_dependencies":true,
        "jupyter_port":9000
    }
}

When the above file is used with the azureml-ngc-tools command, an Azure Machine Learning Compute Cluster named “clara-ct” is created using a node from the “Standard_NC12s_v3” VM size. Follow this link to learn more about the other specifications.

This file lists references to all the NGC Catalog assets that are to be pre-installed in the Azure ML environment to achieve the specific use-case. In this example, the additional resources are the ready-to-use Jupyter notebook and the associated data files. A ready-to-use config file lists the various assets needed for the application below, can be downloaded here. No additional modifications are required for this file.

The ngc_config.json file looks like:

{
    "base_dockerfile":"nvcr.io/nvidia/clara-train-sdk:v3.0",
    "additional_content": {
        "download_content": true,
        "unzip_content": true,
        "upload_content": true,
        "list":[
            {
"url":"https://api.ngc.nvidia.com/v2/models/nvidia/med/clara_train_covid19_3d_ct_classification/versions/1/zip",
                "filename":"clara_train_covid19_3d_ct_classification_1.zip",
                "localdirectory":"clara/experiments/covid19_3d_ct_classification-v2",
                "computedirectory":"clara/experiments/covid19_3d_ct_classification-v2",
                "zipped":true
            },
            {
"url":"https://api.ngc.nvidia.com/v2/resources/nvidia/med/getting_started/versions/1/zip",
                "filename":"clarasdk.zip",
                "localdirectory":"clara",
                "computedirectory":"clara",
                "zipped":true
            },
            {
"url":"https://api.ngc.nvidia.com/v2/resources/nvidia/azuremlclarablogquicklaunch/versions/example/zip",
                "filename":"claractscanexample.zip",
                "localdirectory":"clara/claractscanexample",
                "computedirectory":"clara",
                "zipped":true
            }
        ]
    }
}

When used with the azureml-ngc-tools command, the “nvcr.io/nvidia/clara-train-sdk:v3.0” container is pulled from the NGC Catalog and loaded onto the Azure ML Compute Cluster. Additionally, three resources are downloaded and unzipped into the local environment (with the names “filename” and at the relative directories “localdirectory” provided) and then loaded into the Compute Cluster at the provided location (“computedirectory”).

Run azureml-ngc-tools

Once the two configuration files are ready, run the azureml-ngc-tools command on the local machine to provision the instance:

azureml-ngc-tools --login azure_config.json --app ngc_config.json

Refer to the following screenshot:

Figure 4. AzureML instance being setup by NGC-AzureML Quick Launch Toolkit

The command creates a Compute Cluster “clara-ct” using vmSize :”Standard_NC12s_v3,” then it creates AzureML environment, “clara_env,” with base image “nvcr.io/nvidia/clara-train-sdk:v3.0.” The container is downloaded and subsequently cached after the first use, making it easier for you to reuse the container the next time.

Next, the additional NGC Catalog content is downloaded and unzipped locally, which is then uploaded to the Compute Cluster. The user should be able to see the newly created Compute Cluster, “clara-ct”, in the Azure Portal under the specified workspace, “TutorialTestA1” in this example and tabs Compute ->Compute clusters. This should look like the following:

Figure 5. Console view of AzureML Environment Setup by NGC-AzureML Quick Launch Toolkit

Additional Clara introductory content such as examples of other pre-trained model, datasets and Jupyter notebooks are also downloaded and unzipped locally at the relative directories provided in the “ngc_config.json” file. That content is also uploaded to the Compute Cluster, visible as follows:

Figure 6. Assets from the NGC Catalog pre-loaded onto Azure ML by NGC-AzureML Quick Launch toolkit

The command creates and launches JupyterLab in the Compute Cluster and forwards the port to the local machine so that the Jupyter Lab is accessed on your local machine. A URL is generated that contains the link to JupyterLab running on the Azure ML instance setup, along with the assets specified in the “ngc_app.json” file to be used on your local machine.

Figure 7. Direct URL to fully setup and configured Azure ML JupyterLab for this particular use-case

Once the you have completed your work on the JupyterLab, the Compute Cluster can be stopped by simply entering CTRL+C on the terminal.

The JupyterLab can be launched by copying and pasting the URL into a browser window, as follows:

Figure 8. URL opened on any browser window launches ready-to-use Azure ML JupyterLab

Note that it starts at the workspace root folder for the provided workspace.

To access the content uploaded by the NGC-AzureML Quick Launch Toolkit, you should navigate to the “workspaceblobstore/clara/” folder. All your relevant content will be now available in the session:

Figure 9. Overview of NGC assets pre-loaded onto your AzureML JupyterLab for this use-case

Fine-Tuning a Model with Clara Imaging and Azure  ML

To demonstrate how to build and deploy AI for medical imaging using Clara Imaging and Azure ML, we will fine-tune a pre-trained COVID-19 CT Scan Classification model and optimize it for inference with a custom dataset. This pre-trained model was developed by NVIDIA Clara researchers in collaboration with the NIH, which had a repository of CT radiological images from around the world. The NVIDIA pre-trained model reports an accuracy of over 90% on classification of COVID vs. Non-COVID findings from chest CT scans. More information about the training methodology and results achieved by this pre-trained model are contained in a white paper here.

The model uses two inputs– a CT scan image, and a lung segmentation image. A computerized tomography (CT) scan is a 3D medical imaging procedure that uses computer-processed combinations of many X-ray measurements taken from different angles to produce cross-sectional (tomographic) slices of images. The data needs to be preprocessed, first by converting to Hounsfield units and then rotated to a prescribed orientation, before it can be used for training. A multitude of other pre-trained models developed by NVIDIA Clara for various healthcare applications are made available for free in the NGC Catalog. (N.B. These models are provided for research purposes only.)

The following images show one slice of the stack of images that comprise of the patient’s CT and the corresponding mask:

Figure 10. Example CT-Scan Slice with the corresponding mask (on the right) to focus the model on the lungs

The classifier produces a probability between zero and one, indicating whether the patient has or has not been infected with COVID-19.  However, such predictions require a high degree of accuracy and in most cases, the CT scans used for training may have different characteristics from the original data set used to build the base model. Thus, a model will need to be fine-tuned to achieve the necessary accuracy. NVIDIA Clara includes a fine-tuning mechanism to efficiently adapt the pre-trained model to your dataset, with minimal effort, while increasing the accuracy of your model.

The hospital data for the blog is simulated using 40 labelled data from two sources:

• CT-scans from COVID patients on the COVID-19 CT scans Kaggle Database
• CT-scans from non-COVID CT-scans from The Cancer Imaging Archive.

Running the CovidCT-ScanClassifier.ipynb Jupyter Notebook

Figure 11. The Jupyter notebook for this blog example walks through these steps

Step 1: Set Up

The ready-to-use Jupyter notebook first describes the mechanics of how the pre-trained model is built, and how it can be fine-tuned by introducing the concept of a MMAR (Medical Model ARchive). The dataset from Kaggle is downloaded (by previously obtaining the user’s Kaggle key) and the input examples are examined and visualized. The new data is then indexed in a way that Clara can refer to it for different tasks, such as inferring if the patient has COVID-19 or not, fine tuning, and/or deploying the model. Refer to the CovidCT-ScanClassifier.ipynb Jupyter notebook to see how all these tasks are done with more details.

The data set that we are using to fine-tune the model consists of 40 samples studies. To test the current accuracy of the model, we’ll use three data points from the set. We’ll exclude these three data points when we eventually fine-tune the model.

The Jupyter notebook separates those example as seen here:

Figure 12. Slicing the reference dataset for training and testing

Step 2 Classify/Infer on Test Data

Use the infer.sh command to check the accuracy of the model by using the test data.

The MMAR from the base model has the original configuration files used to train the base model with. Those files need to be adapted to the new data before the infer.sh command is run. The Jupyter notebook will execute all the modifications needed to the required files.

Once the configuration files have been adapted, the infer command is run on the test data.

Inspecting the inference results

The Jupyter Notebook then retrieves the probabilities produced by the “new_infer.sh” command and estimates the predicted labels (1:COVID, 0:NO COVID). Those labels are then used along the true labels to compute the testing examples average precision score using the sklearn.metrics.average_precision_score function.

Figure 13. Testing the pre-trained model for performance without fine-tuning on the new reference dataset

Notice that the average precision (0.833) is not as high as expected (.90 or 90% or higher) because some instances of the new data have either not been preprocessed in Hounsfield units or haven’t be rotated to a specific orientation. We can now improve the accuracy of the model by fine-tuning it with the full data set.

Step 3 Fine Tune         

The train_finetune.sh command executes the fine-tuning mechanism where the original script and its configuration files are adapted to point to the new training data.

Execute new_train_finetune.sh Command

Figure 14. Fine-tuning the Clara COVID-19 CT Scan Classification pre-trained model with training set of reference dataset

Once the training is complete, you can view the training accuracy, training loss, mean accuracy, and the time it took for the fine-tuning to finish. You can decide to either export the model or continue fine-tuning the model. By default, the resulting model is not automatically saved nor exported. The trained model can be exported using the export.sh command. This produces the frozen graphs necessary for the model to be used by other applications, such as Clara Deploy. The Jupyter notebook sets up the new model so that it can be used to classify the test data.

Step 4: Re-test Reclassifying Inference with fine-tuned model

The Jupyter notebook creates a new infer.sh command that points to the full data set and to the new fine-tuned model and then executes the command.

The next step is to compute the Average COVID-19 Classification Precision across all testing examples. The Jupyter Notebook then retrieves the probabilities produced by the “finetuned_infer.sh” command and estimates the predicted labels with 1 (high likelihood of infection) or 0 (low likelihood of infection).

Those labels are then used along the true labels to compute the testing examples’ average precision score using the sklearn.metrics.average_precision_score function.

Figure 15. Testing the fine-tuned model for performance on the testing slice of data from the new reference dataset

Notice that the average precision is now much higher, so the fine-tuning mechanism has succeeded to fine tune the model to account for the peculiarities of the new data.

Deploying the Model

After the AI model is exported and checked, the Medical Model Archive MMAR can then be connected into research workflow pipelines. These pipelines contain operators for the various phases of pre-transforms and inference. The results are consumable by a medical imaging ecosystem. (e.g. a DICOM-SR, a secondary capture image with burnt-in results, an HL7 or FHIR message) The imaging pipeline can be constructed using Clara Deploy SDK, which uses building blocks called operators. Clara Deploy SDK comes with many ready-made pipelines for getting started.

Understanding and mapping the architecture in any given environment is important to know when constructing this pipeline. For example, simulating the connectivity from an image manager like a PACS  by using transformation frameworks to transform the data for the model, is a good first step. Subsequent to the first step, inference tasks follow, and then delivering the outputs in the format accepted by systems and devices at medical institutions.

Figure 16. A sample deployment workflow using the Clara Deploy application framework

When thinking about the results, it is important to consider the types of data being produced and the systems that it will be deployed on. For instance, is it a classification result (presence or absence of a disease, like COVID-19), a DICOM-SR for display within a PACS, or a FHIR Observation object. If it is a segmentation result (identifying a lesion or nodule), creating a DICOM Segmentation object may be appropriate. These are examples of the types of objects consumable by the medical imaging ecosystem and the architecture of the environment is important to know when constructing this pipeline.

Summary

We have shown you how you get started with Clara Train on AzureML for radiological CT images, using a pre-trained AI COVID-19 classification model. This is one example that can be built with the software from the NGC Catalog and AzureML. Beyond just building healthcare centric application, you can use the containers, models, and SDKs from the NGC Catalog to build applications across other use cases, such as conversational AI, recommendation systems, and many more.

Get started today with NVIDIA Clara from the NVIDIA NGC Catalog on Azure ML.

Azure Marketplace new offers – Volume 85

by Scott Muniz | Sep 1, 2020 | Azure, Microsoft, Technology, Uncategorized

This article is contributed. See the original author and article here.

We continue to expand the Azure Marketplace ecosystem. For this volume, 85 new offers successfully met the onboarding criteria and went live. See details of the new offers below:

Applications ABILITY.DatenCockpit: ABILITY.DatenCockpit is a flexible framework for building a customized portal for your employees, customers, or vendors. Combine data from all your organization’s systems and data sources, and tailor dashboards to your needs and working style. APENTO | Consultant Access: This managed service from APENTO grants consultants read-only access to the Microsoft Azure subscriptions of your choice, enabling them to perform tasks or deliver services. Azure Lighthouse delivers an overview of the granted access levels, and you can expand or revoke access at any time. APENTO Managed Services for our Azure customers: In this offering, APENTO will provide your organization with a fully managed enterprise-grade platform powered by Microsoft Azure. APENTO’s highly automated approach can handle even the most demanding applications. APISec: APISec uses smart bots to inspect and map API features, then creates custom security coverage using the Playbook framework (Security as Code) for OWASP top 10, business-logic, and role-configuration flaws. Continuously detect and report vulnerabilities early in the development cycle with APISec. ARCHIBUS on Azure: ARCHIBUS on Microsoft Azure centrally manages an organization’s internal assets and facility-related business information in cooperation with IoT and other service solutions. This app is available only in Japanese.  ArchiveMASH: Bluesource North America delivers ArchiveMASH as part of a fully managed migration service. Bluesource’s dedicated migrations team installs and configures ArchiveMASH, then delivers the migration until all data has reached its destination. Authentication Services: Authentication Services provides enterprise-wide access, authentication, and authorization for UNIX, Linux, and Mac OS X systems by using an organization’s existing Active Directory infrastructure. This allows non-Windows resources to join the Active Directory trusted realm. Azure Sentinel 24×7 Managed Zero Trust Service: Infused Innovations’ threat protection is a managed service that uses your Azure and Microsoft security licensing to deliver a zero-trust environment. Maintain security hygiene on all Windows, Linux, MacOS, iOS, and Android devices and stop zero-day attacks from spreading. Bizdirect CMS – BizOptima: Bizdirect’s cloud optimization services provide you with status updates of your virtual machines, enabling you to identify what is being overused, underused, or unused. Improve your return on investment with Bizdirect. Bromcom MIS: Bromcom is a cloud-based school management information system providing a comprehensive view of pupil and school performance. Bromcom covers attendance, behavior, examinations, assessments, and progress tracking. Features include parent and student portals, online payments, and more. Call Recording Service: Touch Call Recording as a Service enables companies to meet important requirements, including compliance (MiFID, MiFID II, GDPR), documentation, dispute resolution, quality assurance, risk management, audit trails, and training. Boost productivity, add value, and deliver improved customer service. CentOS 7.6 Free: This pre-configured virtual machine from Cognosys launches CentOS, a popular Linux distribution derived from Red Hat Enterprise Linux and used by organizations for development and production servers. CloudTime: CloudTime significantly reduces cloud costs by monitoring your cloud and automatically parking unused resources. Doing so for a just few resources can save thousands of dollars a month on Microsoft Azure bills with little time investment. Connector: Connector helps companies fulfill their obligations with the Brazilian government, both at the municipal level and the federal level. Connector works with NFe, NFSe, and Reinf documents and is available in Portuguese and Spanish. Contactable: PhoneAppli’s phone book and business card management service help manage employee and customer information in the cloud by consolidating all contact information in one place. This app is available only in Japanese. COVID-19 Remote Monitoring & Compliance: ConnectedLife offers a scalable COVID-19 remote monitoring, reporting, and compliance solution that supports the needs of healthcare workers, patients, patients’ families, essential workers, and community volunteers. Cronos Timetable: Using artificial intelligence and mathematical models, Cronos Timetable quickly generates school schedules without teacher and class scheduling conflicts. CyberGate SIP connect to Microsoft Teams: The CyberGate application server lets you connect a SIP H.264 video door intercom to Microsoft Teams. Enable Teams users to answer incoming intercom calls with two-way audio and live video and open the door for visitors. Data360 DQ+: With automated capabilities across data governance, metadata management, data analytics, and data quality, Data360 empowers business and technical users to extract value from data like never before. Data360 makes it easy to identify the trusted data that matters most for your business. Diffusion Real-time Platform: The Diffusion Real-Time API Management Platform delivers centralized data management, optimization, and integration to simplify development, reduce infrastructure requirements, and speed time to market for corporate applications and systems. Digital Patient Experience Manager: Built for the healthcare industry, Digital Patient Experience Manager is a communications automation ecosystem that facilitates automated communications across patients’ preferred channels, including email, SMS, and voice. Automate care reminders and improve patient outcomes. Docker CE with CentOS 8.0: This pre-configured virtual machine from Skylark Cloud launches CentOS, a popular Linux distribution, configured with Docker CE. Docker is a set of platform-as-a-service products that uses OS-level virtualization to deliver software in packages called containers. DRaaS Continuity Planner: Storage Guardian’s free disaster-recovery-as-a-service (DRaaS) platform centralizes the preparation steps to help you plan, test, and communicate your organization’s DRaaS solution. Dynamic Demand Forecast in COVID-19 and New Normal: Singular Intelligence offers its dynamic demand forecast and causal factors-based simulation solution to help users understand how the COVID-19 pandemic has affected the consumer goods and retail industries. Get automated predictive insights, forecasts, alerts, and more. Energy Master Platform: With cost savings being critical for success and energy prices rising, multi-site corporations are challenged to reduce costs and reach sustainability targets. Energy Master Platform helps clients reduce consumption and costs while building strategies around carbon footprint reduction. EntelliFusion: EntelliFusion cleans, shapes, stores, and secures operational data for analytics and machine learning. It is data-agnostic, scalable, and ensures data accuracy and response time performance to promote end-user BI tool adoption. FACILITYtwin: FacilityTwin brings your facility to your fingertips, contextualizing asset maintenance and performance data in mobile and easy-to-use visual interfaces. Transform productivity, risk cost, and flexibility outcomes without interfering with your regulatory processes and systems. Fata Morgana Benchmark: Pocket Virtuality’s Fata Morgana Benchmark offers simple benchmarking for the graphical subsystem of Microsoft HoloLens devices, graphics processing units, and the Fata Morgana rendering engine. FlightPulse: Designed by pilots, FlightPulse is a fully configurable electronic flight bag application that puts key operational data directly in the hand of pilots. Unlock and consolidate disparate data to help pilots make informed, data-driven decisions. GEM Winch by GEM System: GEM Winch is a universal modular system for the transformation, anonymization, and creation of data sets. Easily and intuitively decide which data will be included in a data set, which data to anonymize, and more. This app is available only in Czech. Gorilla Expense: Gorilla Expense provides complete visibility into your organization’s travel and expense spending, and it offers savings through automated expense management. Take the pain out of expense reporting with Gorilla Expense. Industry Weapon MediaBridge for Azure: Developed for Industry Weapon’s CommandCenterHD Platform, the MediaBridge appliance provides a secure outbound connection to the Industry Weapon cloud application and handles the content distribution to digital signage devices in your internal network. Jellyfin – Streaming Media Server on Ubuntu 18.04: This virtual machine image built by Tidal Media Inc. launches Jellyfin Streaming Media Server on Ubuntu. Jellyfin delivers streaming media from your dedicated server to user devices via multiple apps. koel – Music Broadcasting Server on Ubuntu 18.04: This ready-to-run image built by Tidal Media Inc. for web developers contains koel Music Broadcasting Server on Ubuntu 18.04. With a web-based audio streaming service written in Vue on the client side and Laravel on the server side, koel provides flexbox and drag-and-drop functionality. L Squared Digital Signage for Modern Stores: L Squared Hub, a content management system for corporate and retail communication, empowers organizations to deliver high-impact, media-rich content to customers and employees on any digital display. This app is available in English and French. LandView: LandView tracks files associated with oil wells, acreage, payments, obligations, royalties, and division of interest. You can view this data in its reporting tools and title chain visualizer. Leni Pharma Portfolio Management: Leni, an AI-driven solution from LTI, helps portfolio managers in the pharmaceutical industry assess the risk of ongoing clinical trials, simulate the success of future clinical trials, compare against competitor studies, and predict the probability of study success against a set schedule. LTI Enterprise Privacy Rights Management Solution: LTI’s Enterprise Privacy Rights Management Solution (EPRMS) helps organizations manage consumer requests and meet regulatory timelines. It covers consumer consent, identification of personal data, and rules-based anonymization and deletion of customers’ personal data. LTI Mosaic Decisions: Mosaic Decisions, LTI’s data analytics platform, harnesses your valuable data to generate actionable insights, which can then facilitate business transformation and enhance the quality and speed of your decision-making. Managed Detection and Response: The Managed Detection and Response (MDR) service uses Microsoft Azure Sentinel to delegate detection and response to Open Systems AG’s DevSecOps engineers, who are always available to contain threats. Mixed report AI: To reduce the spread of COVID-19, EBILAB’s web app helps restaurants and other retail spaces monitor and predict the number of patrons. This app is available only in Japanese. Neo4j Enterprise 4.0 Causal Cluster: Improve the throughput and processing of your workloads with a Neo4j 4.0 causal cluster. Neo4j is an open-source graph database platform that helps organizations make sense of their data by revealing how people, processes, locations, and systems are interrelated. Neo4j Enterprise VM Version 4.1: Deploy an instance of Neo4j 4.1 on Microsoft Azure. Neo4j 4.1 features more role-based access control and improvements to causal clustering, query planning, and memory management. Neo4j is an open-source native graph database providing ACID-compliant transactions. Optimized CentOS 8 Linux Web Server: This image offered by tunnelbiz.com contains an optimized version of CentOS 8 and PHP 7.4 for improved web server performance. Pocket Work Mate: Companies can reduce turnover of part-time staff with better training and knowledge-sharing resources. Pocket Work Mate helps educate new staff members, automatically generating multilingual subtitles to uploaded videos. This app is available only in Japanese. QR code generation service: Ecoeffy’s QR code app is designed for companies interested in implementing self-service platforms and incorporating barcodes, QR codes, or near-field communication (NFC) readers in their processes. This app is available in English and Spanish. RSA Authentication Manager 8.5.0.0.0: RSA Authentication Manager, the on-premises solution behind the RSA SecurID Access hybrid platform, verifies authentication requests and administers authentication policies for enterprise networks. Simio Manufacturing Process Digital Twin: Design and operate your factory with Simio’s 3D process digital twin. Driven by enterprise data, it provides predictive and prescriptive analysis to improve the design and operations of complex systems. Smart Retail: Control IoT energy costs and reduce your organization’s carbon footprint with Smart Retail. Through remote monitoring, Smart Retail obtains equipment data every 15 minutes, stores it on Microsoft Azure, and processes it with Microsoft Power BI. This app is available only in Spanish. Sproute SPAN SD-WAN gateway: Sproute SPAN SD-WAN gateway helps reduce vendor management burdens; boost network capacity, performance, and availability; improve productivity and security; and enhance visibility throughout your organization. SymbioSys: Distribution Management-as-a-Service: Designed for distributor partner organizations, SymbioSys Distribution Management-as-a-Service helps insurance distributors manage registrations, hierarchies, agent movement, performance, persistency, advisor classifications and rating, incentives, payments, and contests and campaigns. SymbioSys Contest-as-a-Service (for Distributors): Based on consolidated data from all insurers, SymbioSys Contest-as-a-Service helps insurance distributors launch innovative contests to drive sales, promote products, improve revivals, and drive recruitment. SymbioSys Incentive-as-a-Service (for Distributors): SymbioSys Incentive-as-a-Service for insurance distributors uniquely addresses the incentive needs of policy contracts, including life, annuity, pensions, property and casualty, and more. Ensure all your organization’s advisors are well motivated to deliver results with SymbioSys Incentive-as-a-Service. Tick Tock Clock Sync & Monitoring: Tick Tock is a software-based solution for accurate time sync and infrastructure monitoring. It offers sync accuracy as low as 5-to-10 nanoseconds when using hardware timestamps and hundreds of nanoseconds to 5 microseconds when using software timestamps. Tide: Tide is a cloud-based solution that delivers a centralized bordereau transformation, processing, and repository platform. It ensures that delegated authority data is received on time, in a complete, standardized, and contractually compliant format, and is stored in one location. UltraBac Dynamic Replication to Azure: UltraBac’s replication to low-cost Microsoft Azure storage provides organizations of any size with a powerful and simple-to-implement failover protection plan. Continuously replicate on-premises Windows servers to Azure for nearly instant disaster recovery failover. Vuforia Expert Capture: Vuforia Expert Capture is an end-to-end augmented reality solution that drives transformational workforce productivity. It empowers front-line workers to deliver increased business value by always having the relevant procedural guidance to get the job done quickly, accurately, and safely. WANdisco LiveData Plane for ADLS Gen1-Gen2: WANdisco LiveData Plane enables organizations to continue to read and write source data during data migration. Accelerate the move to ADLS Gen2 with no scripting, coding, or business disruption. Wragby Azure Managed Services: Wragby Azure Managed Services provide peace of mind to enable you to focus on business operations while realizing your organization’s digital transformation imperatives. Ynomia Connected Jobsite: Ynomia Connected Jobsite converts worksites into digital warehouses, displaying the real-time location of building materials. Put critical location data in employees’ hands to increase productivity and reduce costs. Zaloni Arena: Arena is an augmented data operations platform providing an active data catalog that enables self-service data enrichment and consumption. Arena is built with an extensible machine learning platform that improves and safeguards enterprise data assets. Zepcam Bodycam Solutions: Zepcam develops body camera solutions for police, law enforcement, military, public safety, public transport, and other industries. Capture video, audio, and GPS information with superior video quality, secure storage, and unique video management features. Consulting services Application Modernization: 2 wk assessment: Dimension Data will assess your applications and dependencies, then provide a report indicating your migration and optimization options and your application readiness to move to Microsoft Azure. Azure Adoption 2-Wk Workshop: In this workshop, Green House Data will cover the Microsoft Cloud Adoption Framework for Azure and create a customized cloud adoption plan for your business. Azure App Services: 4-weeks Implementation Service: With this implementation package, which is intended for organizations facing performance, scale, or security issues with their applications, ComtechRIM India Private Limited will provide assistance in migrating apps to Microsoft Azure App Service. Azure Cost Optimization 3 Week Assessment: MAQ Software’s assessment will help you track your Microsoft Azure resource consumption and identify opportunities to optimize costs, addressing underutilized resources, security, and reporting. Azure Managed Security Services: 2 Wk Assessment: In this engagement, CyberProof, a UST Global company, will conduct a cyber maturity assessment for deploying a security operations center (SOC) platform on your Microsoft Azure environment. Azure Modern Datamart: 1-Day Assessment: Modernize your data mart with an assessment from Intuceo, whose consulting team will analyze your data warehousing goals, propose appropriate Microsoft Azure services, and create a plan to implement a data warehousing pilot. CAF – Strategy, Plan & Ready: 1-Day Workshop: FyrSoft’s workshop will introduce you to the Microsoft Cloud Adoption Framework for Azure, help you put together a cloud adoption plan, and establish or optimize your Azure environment. CenturyLink AVS Migration Assessment 7 Days: This assessment from CenturyLink will provide a high-level indication of costs for an estate on Microsoft Azure VMware Solution, along with recommendations for right-sizing legacy environments ahead of a migration to Azure. Cloud Alignment: 4-hour workshop: Slalom Consulting will meet with your business and technology leaders to determine your organization’s cloud maturity, business and IT alignment, and talent readiness. This workshop is a precursor to Slalom Consulting’s Business Transformation enabled by Cloud (BTeC) engagement. CloudStart CAF Landing Zone: 4-Wk Implementation: ANS Group Limited’s CloudStart framework helps clients quickly configure a production-ready Microsoft Azure environment. ANS Group Limited will guarantee you end-to-end coverage and address any questions you have so you can migrate to Azure with full confidence. Cognitive Services: 2-day Workshop: In this workshop, Awara IT will focus on Microsoft Azure and solutions using cognitive services: voice and image recognition, text translations, smart search, and bots. Only basic Azure skills are required for the workshop, and it’s suitable for any industry and any project scope. Data Exploration: 1-Day Workshop: This workshop from PGS Software will help companies define relevant use cases for machine learning and big data. PGS Software’s data scientists and business analysts will provide guidance on Microsoft Azure data storage solutions, architecture options, and more. Data Lake Starter Kit 1 hour Implementation: In this free engagement, dataroots will implement a module for Terraform, an open-source tool for provisioning and managing cloud infrastructure, that deploys a complete and opinionated data lake network on Microsoft Azure in less than one hour. DevOps on Azure: 2-Week PoC: Infopulse can help you achieve your strategic goals with Microsoft Azure DevOps Services. Infopulse experts will collaborate with you to implement a proof-of-concept project with an appropriate pilot application. Fast & Smart – 2 weeks implementation: 4WARD’s Fast & Smart implementation service makes it easy to switch to Agile work. This offer is available only in Italian and will involve Microsoft Teams, Remote Desktop Gateway, and Windows Virtual Desktop. Industrial IoT: 1 day Assessment: CloudPro’s Microsoft Azure experts will examine your production setup and conduct a short interview with you, then devise a roadmap leading you to Azure IoT and helping you implement industrial IoT in your production environment. Managed IaaS on Azure: 4h Assessment: akquinet AG’s free assessment will take a deeper look into your Microsoft Azure infrastructure and come up with a managed infrastructure offer tailored to your resource needs. Modernise your Data Platform: 2 Hour Briefing: In this free briefing, Data-Driven AI will consider your business goals and requirements, then advise you on how to transition to a modern, advanced analytics platform using Microsoft Azure. Oracle Exadata to Azure 2-Week Assessment US: Data Intensity’s assessment will enable organizations to build solutions on Microsoft Azure for their Oracle workloads. Data Intensity will consider what can be completed with each workload and how to further integrate Azure services. Siebel to Azure – 1-Week Migration Assessment: Is your organization using Siebel for customer relationship management and considering alternatives? Speridian Technologies will analyze your Siebel system and provide a business case for migrating to Microsoft Azure. Snowflake on Azure Consultation – 1 hour briefing: In this briefing, Agile Solutions will describe the advantages of Snowflake when used on Microsoft Azure. Agile Solutions will also cover migration and Azure ETL (extract, transform, load) technology choices. SQL Svr DWH Migration to Azure-10wk Implementation: Brillio will assess business and foundation scenarios to set up a smooth migration from your on-premises SQL Server data warehouse to Microsoft Azure. After the assessment, Brillio will ensure business continuity and governance. Transform & Harden: 6-Wk CAF Azure Implementation: Six Degrees will implement the Microsoft Cloud Adoption Framework for Azure to ensure the right architecture, cost, performance, and security controls are applied to your environment. This will improve efficiency and reduce risk.

Azure Active Directory External Identities goes premium with advanced security for B2C

by Scott Muniz | Sep 1, 2020 | Azure, Microsoft, Technology, Uncategorized

This article is contributed. See the original author and article here.

Howdy folks,

Over the past six months, we have seen organizations adapt to remote business environments and engage with an unprecedented number of external users, and we’ve seen our own service usage growing like crazy for scenarios. With this growth, we also know that security continues to be top of mind.

Today, we are excited to announce the Public Preview of Conditional Access and Identity Protection for Azure Active Directory (Azure AD) B2C. We also have an update to our pricing that makes all Azure AD External Identities features more predictable and affordable with support for premium security features. On the blog today, we’re welcoming back Robin Goldstein, to share more about the new capabilities and pricing.

Regards,

Alex

—–

Hi everyone,

Previously, I shared our Azure AD External Identities vision to make it easier to secure, manage and build apps for collaborating and connecting with external users. Today’s announcement, which adds Conditional Access and Identity Protection to Azure AD B2C, is another step in our journey to help organizations protect all their identities with Microsoft’s world-class security features.

Protect your customers, apps, and brand 

Security is essential to maintaining your customers’ and consumers’ trust and protecting your organization’s data. That’s why so many of our Azure AD customers rely on Azure AD Identity Protection and Azure AD Conditional Access. Identity Protection helps organizations automatically protect against account compromise with cloud intelligence, powered by advanced risk detection based on heuristics, User and Entity Behavior Analytics (UEBA), and machine learning (ML) from signals across the Microsoft ecosystem. By pairing Conditional Access policies with the power of Identity Protection, admins can. This combination already helps Fortune 500 organizations prevent over 10 million attacks each month.

By making risk-based Conditional Access and risk detection features of Identity Protection available in Azure AD B2C, our customers can now:

• Leverage intelligent insights to assess risk with B2C apps and end user accounts. Powered by signals from billions of monthly authentications across Azure AD and Microsoft accounts, Identity Protection real-time ML algorithms use adaptive intelligence to flag authentications as low, medium, or high risk and gets smarter and more accurate over time. Detections include atypical travel, anonymous IP addresses, malware-linked IP addresses, and Azure AD threat intelligence. Portal and API-based reports are also available.

Examine risk detections flagged through the portal

• Automatically address risks by configuring adaptive authentication policies for B2C users. By tailoring Conditional Access policies, app developers and administrators can mitigate real-time risk by requiring multi-factor authentication (MFA) or blocking access depending on the user risk level detected, with additional controls available based on location, group, and app.

Create a Conditional Access policy based on the level of sign-in risk

• Integrate with Azure AD B2C user flows and custom policies. Conditions can be triggered from built-in user flows in Azure AD B2C or can be incorporated into B2C custom policies. As with other aspects of the B2C user flow, end user experience messaging can be customized according to the organization’s voice, brand, and mitigation alternatives.

Create a user flow in Azure AD B2C

Get started today

Conditional Access and Identity Protection for Azure Active Directory (Azure AD) B2C will be progressively rolled out across Azure regions starting on September 1. To start using the new features, you will need to do a few things first:

1. Link your Azure AD B2C tenant to an Azure subscription. This allows you to use Azure AD premium features for Azure AD External Identities and Azure AD B2C.
2. Enable Premium P2 features in your Azure AD B2C tenant. This allows you to access Identity Protection and all future Premium P2 features for Azure AD External Identities and Azure AD B2C.
3. Start using the new recommended user flows in Azure AD B2C. This allows you to connect to your Conditional Access policies and all new features added to B2C user flows in the future.

Selecting and creating recommended user flows in Azure AD B2C

And of course, you can always check out our documentation to learn more about setting up risk-based Conditional Access policies for Azure AD B2C.

Pricing update

Azure AD External Identities features are now available as part of one unified offer based on monthly active usage. Whether you use Azure AD B2C, B2B collaboration or the new self-service sign-up features in Azure AD, securing and managing external users is more affordable than ever, with the first 50,000 monthly active users free at both the Premium P1 and Premium P2 tiers. For more information on the new offer with pricing details click For documentation to set up your tenants for this offer, click here.

As always, we hope you’ll try out the new features and share feedback through the Azure forum or by following @AzureAD on Twitter. 

Thanks!

Robin

Learn more about Microsoft identity:

• Related Articles: Learn more about the Azure AD External Identities features announced at Build 2020 and public preview of web API integrations
• Return to the Azure Active Directory Identity blog home
• Join the conversation on Twitter and LinkedIn
• Share product suggestions on the Azure Feedback Forum