Build and deploy the MyDriving solution to your environment
MyDriving is an Internet of Things (IoT) solution that gathers data from your car, processes it by using machine learning, and presents it on your mobile phone. The back end consists of a variety of services provided by Microsoft Azure. The clients can be Android, iOS, or Windows 10 phones.
We created the MyDriving solution to give you a jumpstart in creating your own IoT system. From the MyDriving repository on GitHub, you can get Azure Resource Manager scripts to deploy the back-end architecture into your own Azure account. From that point, you can reconfigure the different services, modify the queries to suit your own data, and so on. You can find these scripts--along with code for the mobile app, the Azure App Service API project, and more--in the MyDriving repository.
If you haven't tried the app yet, look at the Get started guide.
There's a detailed account of the architecture in the MyDriving Reference Guide. In summary, there are several pieces that we set up to create a similar project:
- A client app runs on Android, iOS, and Windows 10 phones. We use the Xamarin platform to share much of the code, which is stored on GitHub under
src/MobileApp. The app actually performs two distinct functions:
- It relays telemetry from the on-board diagnostics (OBD) device and from its own location service to the system's cloud back end.
- It's a user interface where users can query about their recorded road trips.
- A cloud service ingests the road-trip data in real time and processes it. The main work of creating this service is to choose, parameterize, and wire up a variety of Azure services. Some of the parts require scripts to filter and process the incoming data. We use an Azure Resource Manager template to configure all the parts.
- A mobile service app is the web service behind the user interface part of the device app. Its main job is to query the database of stored, processed data. Its code is on GitHub under
- Visual Studio with Xamarin is our development environment. Xamarin, which exists both as a component of Visual Studio and as a stand-alone integrated development environment (IDE), is used to build the cross-platform device code. To build the iOS code, it's necessary to have an instance of Xamarin running on an OS X machine. If required, it can be run as an agent, managed from Visual Studio.
- Unit testing of the device apps is performed in Xamarin Test Cloud.
- GitHub is the repository where we store all the code, scripts, and templates.
- Visual Studio Team Services is a cloud service that's used to manage the continuous build and test of the web service and device apps.
- HockeyApp is used to distribute releases of the device code. It also collects crash and usage reports and user feedback.
- Visual Studio Application Insights monitors the mobile web service.
So, let's see how we set up all of that.
Note: Many of the following steps are optional.
Sign up for accounts
- Visual Studio Dev Essentials. This free program provides easy access to many developer tools and services, including Visual Studio, Visual Studio Team Services, and Azure. It gives you a $25/month credit on Azure for 12 months. It also includes subscriptions to Pluralsight training and Xamarin University. You can also sign up separately for free tiers of Azure and Visual Studio Team Services, but these do not provide Azure credits.
- HockeyApp (optional), for managing test distribution of mobile apps and collecting telemetry.
- Xamarin (required), for building the mobile app and running debug runs and tests on Xamarin Test Cloud.
- GitHub (optional), to create free public repositories for your own code (private repositories are paid). Alternatively, you can use the basic plan in Visual Studio Team Services for private repositories.
- Power BI (optional), to create rich visualizations of data across the entire system.
Note: You don't need a GitHub account to access the MyDriving code in the GitHub MyDriving repository.
Install development tools
The following setup is for developing the full solution: an iOS, Android, and Windows 10 Mobile cross-platform app, with an Azure back end.
As an alternative, you can use Xamarin Studio on either Mac or Windows to develop the mobile apps if you aren't working on the Azure back end.
There's a longer description of this setup.
Windows development machine
The central tool on Windows is Visual Studio, for working with the MyDriving app for Android and Windows, the App Service API project, and microservice extensions.
Xamarin, Git, emulators, and other useful components are all integrated with Visual Studio.
- Visual Studio with Xamarin (any edition--Community is free).
- SQLite for Universal Windows Platform. Required to build the Windows 10 Mobile code.
- Azure SDK for Visual Studio. Gives you the SDK for running apps in Azure, along with command-line tools for managing Azure.
- Azure Service Fabric SDK. Required to build the microservice extension.
Be sure you have the right Visual Studio extensions. Check that under Tools, you see Android, iOS, Xamarin…. If not, open Visual Studio, search for Xamarin, and follow the prompts to install it. Also, check that Git for Windows is installed. If not, in Visual Studio, search for it and follow the prompts to install it.
Mac development machine
The Mac (Yosemite or later) is required if you want to develop for iOS. Although we use Visual Studio with Xamarin on Windows to develop and manage all the code, Xamarin uses an agent installed on a Mac in order to build and sign the iOS code.
(As an alternative, you can use Xamarin Studio directly on the Mac to develop cross-platform apps.)
You don't need the Mac if you don't want to include iOS as a target platform.
- Xamarin Studio for iOS. You can also set up Visual Studio and Xamarin on a Mac that's running a Windows virtual machine. See Setup, install, and verifications for Mac users on MSDN.
- Azure development tools (optional).
Enable remote login on the Mac. Open System Preferences > Sharing, and then select Remote Login.
When you open an iOS project in Visual Studio on Windows, the Xamarin plug-in will prompt you for the ID of the Mac.
Fetch the GitHub repository
Fetch a local copy of the GitHub MyDriving repository by using the Download ZIP button in GitHub, Visual Studio, or another Git client.
Unzip the file to a folder with a short path name, such as C:\code.
Alternatively, if you want to keep up to date with or contribute to our code, clone the repository as follows:
Get a Bing maps API key
You need to replace this in line 22 in
Build the demo app
Open these solutions in Visual Studio:
You'll get prompts to:
- Trust some potentially untrustworthy projects. Choose to open them if you want to go ahead.
- Set Developer mode if you're working on a fresh Windows 10 machine.
- Provide your Xamarin credentials.
- Connect to the Xamarin Mac. If you don't have a Mac, right-click the iOS project in Visual Studio, and then select Unload project.
Rebuild the solution.
If you have trouble building, try the solutions to quirks that we've found:
VINLookupApplication project doesn't load: Be sure that you installed the Azure SDK for Visual Studio.
Service Fabric project doesn't build: Build the interface projects first, and be sure that you installed the Service Fabric SDK.
Android app doesn't build:
- Open Tools > Android > Android SDK Manager, and ensure that Android 6 (API 23)/SDK Platform is installed.
- Delete this directory, and then rebuild:
Get to know the code
In the solution, you'll find:
Azure extensions: Service Fabric.
Azure HDInsight: Scripts for processing trip data in Azure.
Mobile Apps: The device apps.
MobileAppsService/MyDrivingService: The web back end.
Power BI: The report definition.
- Resource Manager: Templates to build the Azure resources.
- PowerShell: Scripts to run the Resource Manager templates.
- Azure SQL Database: Debugging databases.
SQL Database: CreateTables: schema definitions.
Azure Stream Analytics: Queries that transform the incoming data stream.
Run the apps in development mode
Take action to run the apps, based on the device that you're using:
Back end: Set MyDrivingService as the startup project, and press F5 to run the back-end web service. It will open a browser view of the API listing.
Mobile clients: The mobile apps are developed in Xamarin.
Upload the mobile app to HockeyApp
HockeyApp manages the distribution of your Android, iOS, or Windows app to test users, notifying users of new releases. It also collects useful crash reports, user feedback with screenshots, and usage metrics.
Start by uploading your build app. Then, sign in to HockeyApp from your development machine. On the developer dashboard, click New App, and then drag the built files onto the window. (Later, you can automate your build service to do this.)
Now, you're in your app dashboard.
Repeat the process for each platform that your app runs on. Then, you can do the following:
Use the app ID from the dashboard to send crash data and feedback from your app. In MyDriving, update the IDs in src/MobileApps/MyDriving/MyDriving.Utils/Logger.cs.
Invite test users. You get a URL to recruit testers users. They'll be able to sign up for your team, download the app, and send you feedback.
If you'd prefer a more open beta release, set the distribution to public. Click Manage App > Distribution > Download = Public. Now anyone can download your app and send you feedback, and they'll see a notification when you post a new version. You might get some crash reports from them too.
Link crash reports to Visual Studio Team Services. Click Manage App > Visual Studio Team Services. HockeyApp can automatically create work items in Team Services when there are crash reports or when feedback is received.
Read more at the HockeyApp site.
Test the mobile app on Xamarin Test Cloud
Xamarin Test Cloud automates UI testing on real devices in the cloud. By using the NUnit framework, you write tests that run your app through the user interface.
To use Xamarin, you incorporate the Xamarin.UITests SDK into your app, which comes as a NuGet package. You'll find it in the demo app, and it's included when you create new test projects with the Xamarin templates.
If you use a Visual Studio Team Services build, it's easy to write Xamarin UI unit tests and run them as part of your build.
Deploy Azure services
To perform an automatic deployment of Azure services and Team Services build services, refer to the detailed instructions in scripts/README.md.
Microsoft Azure provides a wealth of different services that you can use to build cloud applications. Although many can be used individually (such as App Service/Web Apps), they're at their best when they're interconnected to form an integrated system like that we use in MyDriving.
It's possible to create and interconnect Azure services manually, but it's much quicker and more reliable to use Azure Resource Manager templates. Resource Manager automates the deployment of a solution's resources and making the interconnections between them.
You'll find the template for the MyDriving system in the GitHub repository under scripts/ARM. It provides a comprehensive and concise view of how the different services in our architecture are interconnected. We explain all these in detail in the MyDriving Reference Guide, but you can learn a lot just by reading through the template itself.
Note: Most Azure services have an associated cost, depending on the pricing tier. If you're new to Azure, you can try it out for free. However, if you don't plan to use certain components in the MyDriving system, be sure to remove them to avoid incurring costs. The "Estimate operational costs" section later in this article provides a summary of typical service expenses.
Edit the template
To customize your deployment, perhaps to remove unneeded components or to add others, first make a copies of scenario_complete.params.json and scenario_complete.json in which to make changes.
You can use the scenario_complete.params.json file to override various default values, such as the service SKU or the storage replication type, as described in the following table. The default values select the lowest-cost options.
|IoT Hub SKU||Tier for Azure IoT Hub service||F1|
|Storage Account Type||Storage replication type||Standard LRS|
|SQL Service Objective||Concurrency slot consumption||DW100|
|Hosting Plan SKU||Service plan for App Service||F1|
- Search for "baseName" and change it to a name that you prefer.
- Search for "Create". Each of these sections creates a resource.
- Set sqlServerAdminLogin and sqlServerAdminPassword to suitable values.
- Before you delete a section that creates a resource, check whether it has dependents by searching for its name elsewhere in the file. Note that each section that creates a service includes a dependsOn section that lists its dependencies.
Here's what the template configures. Details are in the Reference Guide.
|Service||Description and details|
|Storage accounts||The template creates three accounts:|
|- A SQL database that receives aggregated telemetry from Stream Analytics, and serves as the backing store for Azure App Service tables that expose this data through API endpoints.|
|- Blob storage that accumulates historical data from another Stream Analytics job, to be processed by HDInsight.|
|- A SQL database that receives results processed by HDInsight for use with Power BI.|
|Azure IoT Hub||Establishes a two-way connection to each connected device. In the MyDriving solution, the mobile app acts as a field gateway to send data to Azure IoT Hub. Azure IoT Hub then serves as an input to Stream Analytics.|
|Azure Event Hubs||An output for a Stream Analytics job that queues the output to extensions that are created with Azure Service Fabric.|
|Azure SQL Data Warehouse|
|Stream Analytics jobs||Connect inputs and outputs with a query, which is used to aggregate both real-time and historical data for the App Service APIs, Azure Machine Learning, extensions, and Power BI.|
|Machine Learning workspace||Includes experiments, R code, and API service.|
|Azure Data Factory||Scheduled Machine Learning retraining.|
|Service Fabric hosting plan||For extensions.|
|App Service (“Mobile App”)||Hosts the Mobile Apps API project that provides endpoints for the mobile app. The API code must be deployed to App Service from Visual Studio.|
|Alert rules||Sends you email if the app responses indicate failures.|
|Application Insights||For monitoring performance of the APIs in App Service. You have to configure the connection in Visual Studio.|
|Azure Key Vault||For saving the web service cluster certificate.|
Run the template
In scripts/README.md, there are detailed instructions for running the template.
To provision all these services in your own Azure account by using the script, do one of the following:
cd scripts/PowerShell; deploy.ps1 *location* *resourceGroupName*
- location is the Azure location, such as
West US. Use
Get-AzureLocationto find a list of available locations.
- resourceGroupName is the name that you want to give to the group that all the resources will belong to. When you're finished with the resources, you can delete them all together by deleting this group.
- location is the Azure location, such as
Run DeploymentScripts/Bash/deploy.sh with Bash.
Open and build the Visual Studio solution DeploymentScripts/VS/DeployARM.sln.
Note that each time the template is run, it creates a new set of resources with new names. To delete the resources, go to the portal and delete the resource group.
If the script fails for any reason, you can re-run it.
The script gives you the option of configuring continuous integration in Visual Studio Team Services. If you have set up a Team Services project, you'll have a URL: https://yourAccountName.visualstudio.com. Enter the complete URL when you're asked. You can give it a new or existing name for a Team Services project.
Set up build and test definitions in Visual Studio Team Services
We use Team Services on this project mostly for its build and test features. But, it also provides excellent collaboration support, such as task management with Kanban boards, code review integrated with tasks and source control, and gated builds. It integrates well with other tools such as GitHub, Xamarin, HockeyApp, and of course, Visual Studio. You can access it through the web interface or through Visual Studio, whichever is more convenient at any moment.
The steps in the build and release definitions use a variety of plug-in services that are available in the Team Services Marketplace. In addition to basic utilities to run command lines or copy files, there are services that invoke builds by Xamarin, Android, and other vendors, and that connect to HockeyApp.
We have build definitions for each of the main targets. We also have variations for feature and regression testing. That gives us:
MyDriving.Services (the back-end web app for the mobile app)
If you want to see the full details of our configuration, see section 4.7 of the MyDriving Reference Guide, "Build and Release Configuration." They follow the same general pattern. The script:
- Restores the NuGet package. We don't keep compiled code in the repository, so the first steps of each build are to restore the required NuGet packages.
- Activates the license. The build is performed in the cloud, so where we need a license--in particular, for the Xamarin build service--we have to activate our license on the current build machine. Then, we deactivate it immediately afterward, to allow it to be used on another machine.
- Builds by using the appropriate service. We use Xamarin builds for the mobile apps, and Visual Studio builds for the back-end web service.
- Builds tests.
- Runs tests. We run the mobile app tests in Xamarin Test Cloud.
- Publishes the build result to the drop location.
The trigger for the main builds is set to continuous integration. That is, the build is run every time code is checked in to the master branch.
Release definitions are set up in much the same way.
For the web service, we set up deployment as an Azure web app:
And we set the release trigger to continuous deployment. That is, every check-in followed by a successful build results in an update to the web app.
For mobile apps, we deploy to HockeyApp:
Explore telemetry by using Application Insights
Application Insights collects telemetry about the performance and usage of your web services. The Application Insights SDK sends telemetry from the service to the Application Insights resource in Azure.
Browse to the Application Insights resource that the template set up. There, you can explore charts of the performance of your Mobile App Service project. They show server requests and response times, failures, and exception counts. There are also charts of dependency response times--that is, calls to the database and to REST APIs such as Machine Learning. If there are any performance problems, you'll be able to see what piece of your system is causing them.
If you have a web service that you set up by hand, it's easy to get the same charts. On the web service blade, click Tools > Extensions > Add. Select Application Insights.
The feature works by instrumenting your application with the Application Insights SDK.
You can add custom telemetry (or instrument an application that's running somewhere outside Azure) by adding the Application Insights SDK at development time. This is useful to log metrics that depend on the application, such as users' average trip length or total mileage. In Visual Studio, right-click the project, and then select Add Application Insights.
Application Insights will send alert emails if it sees unusual numbers of failure responses. You can also set up your own alerts on various metrics, such as response times.
Just to be sure that your web service is always up and running, you can set up availability tests. These tests ping your site from various locations around the world every 15 minutes. Again, you'll get an email if there seems to be a problem.
Estimate operational costs
It's remarkably inexpensive to run an app like this one at small scale. Many of the services have free entry-level tiers, so that development and small-scale operation cost very little. And of course, your own apps don't have to use all the features demonstrated in MyDriving.
Here's a rough estimate of our costs in setting up the development configuration for MyDriving. We also note some alternatives that we did not use. This information might be helpful as you estimate your own costs.
- A team of no more than five (plus observing stakeholders).
- Running for about a month.
- 100 users with four trips per day.
Note: If you're new to Azure, there's a free account.
|Visual Studio 2015 Community with Xamarin
Cross-platform dev environment
|Visual Studio Community. (Need Visual Studio Professional for Xamarin.Forms, to design cross-platform from a single code base.)||$0|
|Azure IoT Hub
Two-way data connection to devices
|8,000 messages + 0.5 KB/message free.||$0|
High-volume stream data processing
|Charge of $0.031 per streaming unit per hour, while enabled. You choose the number of streaming units that you want; more to scale up.||$23|
+ 3-hour experiment * $1 / experiment hour.
+ 3.5-hour API CPU * $2 / production CPU hour.
API CPU time assumes 5 min/day retraining, though this would rise with more input data.
+ 2 min/day scoring to process 400 trips/day.
Host for mobile back end
|Tier B1--production web apps.||$56|
|Visual Studio Team Services
Build, unit test, and release management; task management
|Private agents, five users.||$0|
Monitoring of performance and usage of web services and sites
Distribution of beta apps, plus collection of feedback, usage, and crash data
|Two free apps for new users.
Code on a uniform platform for multiple devices
|SQL Database for Azure App Service||Basic tier; single database model.||$5|
|Service Fabric (optional)||Run a local cluster.||$0|
Versatile displays and investigation of streamed and static data
|Free tier: 1 GB, 10,000 rows/hour, daily refresh.
$10/user/month for higher limits, more connection options, collaboration.
|Storage||L (locally redundant) < 100 G $0.024/GB.||$3|
|Data Factory||$0.60 per activity * (8 - 5 FOC).||$2|
On-demand cluster for daily retraining
|Three A3 nodes at $0.32/hour for 1 hour daily * 31 days.||$30|
|Event Hubs||Basic with $11/month throughput unit + $0.028 ingress.||$11|
For more information, see:
Send us your feedback
Because we created MyDriving to help jumpstart your own IoT systems, we certainly want to hear from you about how well it works. Let us know if:
- You run into difficulties or challenges.
- There is an extension point that would make it more suitable to your scenario.
- You find a more efficient way to accomplish certain needs.
- You have any other suggestions for improving MyDriving or this documentation.
To give feedback, file an [issue on GitHub] or leave a comment below (en-us edition).
We look forward to hearing from you!
We recommend the MyDriving Reference Guide, which is a comprehensive description of the design of the system and its components.