Developments in Microsoft Azure: report-out from the Azure Red Shirt Dev Tour

Azure is Microsoft’s platform for cloud services and cloud computing and is a core part of Microsoft’s vision for the future. It gets a great deal of attention at Microsoft internally, which isn’t surprising if you consider that Azure sales more than doubled in recent quarters. Just to give you an idea: in Q1 of 2017 the Azure business grew by 93%. The total annual run rate of Microsoft’s commercial cloud business (which includes Office 365) now exceeds $14 billion and is expected to hit $20 billion by the year 2018.
Another illustration of Microsoft’s commitment to Azure is the recent Azure Red Shirt Dev Tour (there’s a short and rather funny introduction video of the event on YouTube), in which long-time Microsoft evangelist Scott Guthrie, one of Microsoft’s top speakers for the developer community, toured Europe for day-long presentations about all things Azure, hitting cities like London, Dublin, Oslo and Amsterdam. Registration was free of charge, and the Amsterdam venue that I attended was fully booked with around 1200 attendees.


Guthrie, famous for his red polo shirt (hence the “Red Shirt” tour – it would have been interesting if he had worn a shirt combining red and azure for the occasion, covered a number of capabilities recently added to Azure in an impressive three hour keynote full of hands-on live demos, followed by an afternoon session of tips & tricks. It was a good showcase of the many capabilities any cloud platform has to offer nowadays, and a glimpse of what is coming in the near future. These are some of the highlights:

A different VM type for each purpose

Probably the most used feature in Azure is still the capability to run a Virtual Machine in the cloud. There is a considerable number of standard VM types suited for different purposes, going from basic low-cost VM’s to high-performance monsters. Some of the latest additions are VMs that include a GPU, and a very impressive VM type that was created specifically to host big SAP databases (like SAP HANA) on Azure. This new M-series VM has some impressive specs, running up to 128 CPU cores with 3.5 TB of RAM (going up to 20 TB RAM is also an option).

128 CPU cores in action in a single M-series VM

The more powerful VM’s can also run Hyper-V, which enables them to run other VMs inside the main VM. And since things like a Linux VM are also an option in Azure, you can now for example run a Windows Server VM in the cloud which in turn hosts a number Linux VM’s.
Another new feature is an interesting way to help customers save money on licenses (which is at the same time an incentive for customers to not hesitate about creating more VM’s of course): on the VM creation screen there is now a “Save money” option at the bottom. If a customer already has a Windows Server license with Software Assurance, you are entitled to a 40% price reduction as part of the ‘Hybrid Use Benefit’ that’s part of the Software Assurance.


A striking element of nearly each demo in this event was the omni-presence of non-Microsoft technology: several demos were given from an iPhone or a MacBook using a Chrome browser, there was lots of open-source scripting, Linux, and databases other than SQL Server (like MySQL, Progress).
Also noteworthy was the complete absence of Windows Phones – the same thing happened at the recent Microsoft Build conference. I’m clearly not the only one who recently abandoned the Windows Phone platform (see my earlier blog post). Even Microsoft employees now openly use Android or iOS phones.

Azure App development and monitoring

Microsoft’s development tooling has always been a very strong part of its technology stack. The integration with Azure is also there. If there is a failure inside an Azure app, you can see it from the App monitoring functionality. Then, from the detail screen for that failure, you can directly create a new work item inside Visual Studio Team Services (VSTS).

A money-saver is the auto-scale option for Web Apps in Azure. This allows an admin to define rules that say something like:  if CPU usage is more than 70% for x amount of time, then scale up.
An option often overlooked is that you can of course also scale down (below your usual minimal spec) at the hours of the day when your app is hardly being used.

Another demo showed the option to do continuous delivery: update the code for an Azure App in GitHub, which automatically triggers a build, test and (optionally) deploy cycle. This kind of Continuous Integration & Continuous Delivery (CI/CD) pipeline can be configured from Azure portal.
Note that the build servers used for this – which you previously had to set up yourself – are included as part of the service. Azure handles this internally, so you do need to spend any time on the build server infrastructure.
Deploying apps in self-supporting containers is another fast-rising trend. Visual Studio 2017 now has integrated Docker support and tooling. You can even include ‘dockerization’ as part of your CI/CD pipeline.

Deployment slots are also a powerful feature, which allows you to deploy the new version of your app to an alternative URL so you can check the result. This second location functions as a staging slot. At a time of your choosing you can then simply swap the production slot with the staging slot. The production URL now has the app version that was in the staging slot and vice versa.

Native mobile apps

Xamarin, the development tool to create native mobile apps (purchased by Microsoft a few years back) has now been fully integrated into Visual Studio 2017. It’s also suited for iOS app development, you do not even need a MacBook for proper testing, Xamarin’s Live Player allows you to view how your app will look on a certain device. In the Azure mobile center you can test the deployment of your app on actual devices that are racked up in the cloud. This is especially useful for Android devices since there are so many different Android versions.


There are of course also important developments for databases.  SQL Server can be run as a VM but also as a service: no machine to log onto, you just get the connection string to your database. A large instance of such a database can now hold up to 4 TB of data.
Next to SQL Server, you can now also get PostgreSQL and MySQL as a service.

Microsoft recently announced the Azure Database Migration Service, which provides a Lift & Shift service for any current (on-premises) SQL Server database to the Azure database service. The migration service promises that there will no (or hardly any) need to change any code related to your database.

Azure also contains a Performance Recommendation option, which works based on machine learning. It is an easy way to tune your database. According to Guthrie, proper tuning can often prevent the need for a larger (and more expensive) database.

If you want to go beyond relational data, and store any kind of data (like key-data, documents, graphs) on a global scale, Azure now offers Cosmos DB (formerly known as DocumentDB). This is a rather impressive, globally distributed database service. In a nutshell: any data you want, synchronized across any region in the world where you need it.
Cosmos DB supports many kinds of data and you can use multiple open source API’s against it. It can scale to millions of transactions p/sec and petabytes of data. The Cosmos DB service offers a comprehensive SLA on four dimensions: there are 99.99% guarantees for availability, throughput, latency, and consistency.

Serverless computing

Why pay for a VM or even a service that is always on, when you actually just need a certain function when a customer asks for it? One of the biggest trends for the coming years is serverless computing, allowing you to trigger a piece of serverless code only when you need it, paying for only the clock cycles your are actually using. This dramatically lowers the cost to only a fraction of an average Azure Compute scenario.
The serverless code can be defined as part of Azure Functions, or you can define a serverless workflow as part of an Azure Logic App. There is already a large number of predefined event sources available that can trigger an Azure Function or a step in an Azure Logic App.

Cognitive Services

The Azure Cognitive Services are accessible quite easily through a proper API. An app was shown for example where you could upload cat images. The cognitive service determined on the fly if the image actually contained a cat, otherwise the image was rejected and moved to a separate folder.
A final demo was the Kiosk Realtime Crowd Insight (this demo is available on GitHub) that shows the estimated age, gender and mood of the person in front of the webcam.
We talk a lot about Machine Learning these days, but I still found it almost staggering how easy it has become to apply this kind of functionality in your own application at very low cost.

This article was originally posted on the Capgemini ‘Capping IT Off’ blog.


Why I abandoned the best mobile operating system

This is a story about how I abandoned my old mobile phone last year, even though it had the best mobile user experience available. Or actually, it’s a story about the digital workplace. Because let’s face it: considering the amount of time we spend on our phones, the smartphone is an important part of that digital workplace.

So what phone did I have you ask? It was a Windows Phone (yes, you can stop snickering now), the Lumia 830 if you want specifics; the last one branded as Nokia. It runs the Windows Phone mobile client OS that was originally released in 2010, as follow-up to Windows Mobile (formerly PocketPC). While Windows Mobile was very Microsoft-oriented in terms of connections and user interface, the new Windows Phone OS was developed from the ground up with the capabilities of modern smartphones in mind and a more universal (think cross-platform) look onto the world. This new OS arrived very late in the game, because competitors Android and iOS already had well developed operating systems for modern smartphones at the time. The advantage of arriving late though, was that Microsoft had a chance to take all lessons learned from existing mobile operating systems, and develop a whole new user interface taking all this into account.

I have always found the iOS user interface to be cluttered, with a plethora of different looking icons on your start screen. The Android interface has the same issue and was still rather clumsy around 2010 (although much improved since then). In comparison, the Windows Phone’s interface of tiles is clean, structured, intuitive, and you can get to any setting or application really fast. Also the so-called live-tiles are far superior to any widget I’ve seen on an Android phone, because of the way they fit and blend into the overall appearance. Add to that the flawless Exchange, OneDrive and Skype integration, and you get an almost unbeatable mobile digital workplace in terms of productivity.


Why abandon my phone then? That had everything to do with available apps and app support. So how did this happen? Well, do you remember the VHS vs. Betamax format war in the 80s? Or more recently HD-DVD vs. Blu-ray? SD-card vs. Memory Stick? Super Audio CD vs. regular CD? My point is: the best platform doesn’t always win. If another product gets better marketing, a better image, better distribution channels, more resellers, then an objective comparison of features and usability doesn’t have to be all-decisive. And at some point if you drop below a certain market share, the battle is lost. This is exactly what happened with the Windows Phone – for a large part I should say because Microsoft was so late in the game, and the competition had already cornered the market.

As a result nearly every company or event organizer that was building a new mobile app, made the decision to only create a version for Android and iOS, because the limited Windows Phone user base didn’t justify the cost of an additional version. There still were many Windows Phone apps of course for key applications (mobile banking, Facebook, Twitter), sometimes sponsored or even created by Microsoft, but often these were updated less frequently and had less capabilities than their (better looking) Android/iOS versions. And then over the last few years, when Microsoft basically stopped investing in phones and the Windows Phone OS altogether, focusing more on Windows 10 and the Universal Windows Platform Apps, many companies even dropped support for their existing Windows Phone apps. It was a hopeless downward spiral.

So what does this tell us about digital customer experience in general? Most of all, although user experience is important, it’s not the whole story. You may have an app or a site with a slick look & feel and the best usability around, but that’s just the storefront. Important as that is, if you don’t back it up with proper marketing, image-building, good distribution channels and continuous updates, you are unlikely to succeed. In a market that is becoming more and more digital and global, keeping both the storefront and all surrounding processes in shape takes considerable effort and attention.
And as the Windows Phone example shows us, you also need to act in time. A great digital user experience is important, but that by itself is not the killer feature that will lure customers away from one of your competitors. Gaining market share (or preventing market share loss) if you arrive late in the digital game is going to be tough.

This post was originally published on the Capgemini ‘Capping IT Off’ blog.

Versioning in SharePoint when editing in Word compared to Word Online

While I was developing a workflow to be triggered on a changed Document Library item in SharePoint Online (this was a Word document), I noticed something interesting about the versioning.
In the workflow I was checking if the version number had already increased, thinking that in this way I could determine if the user was done editing.

As it turns out, it depends if you are using Word or Word Online. What happens is this:
If you are using your local MS Word installation, then you are basically working offline. You only connect to SharePoint Online when you take an explicit action like saving or checking-in a document. SharePoint will not create a new version in the version history until you click Save in Word. So even if you type lots of new text and update some properties, but close the document without saving, nothing is changed in the version history.

This is very different in Word Online. As you probably know, there is no Save button in Word Online, because any change is automatically saved for you. So that made me wonder: if I edit a document in Word Online for half an hour, making many changes, then that document is going to be saved automatically dozens of times. So does that create dozens of new versions in the history? Well, that would be awkward and luckily the answer is no.

What happens is that Word Online creates a new version in the version history (in SharePoint Online) when the first change is being saved in Word Online. So as long as you don’t touch anything in Word Online and just read the document, you get no new version. Then, on every next change in the document that is saved in Word Online, that same new version is *updated*. You can see the Modified time change in the version history.
That is actually the same behavior that we are used to in (local) Word: as long as you do not close the document, subsequent saves will be considered as an update to the same version. This is true if you use both minor and major versions, but also if you only use major versions.

So coming back to my workflow, the downside here is that a new version in the history does not always mean that the user explicitly saved his work. That depends if Word Online or (local) Word was used. Your new version will in any case appear in your version history more quickly when using Word Online.


Search results not showing duplicates

Recently I ran into the issue that the SharePoint Online search was not returning all expected search results – or so it seemed.

To do a test on a site with many items, I uploaded around 7000 documents, all with a different file name but the contents were identical.
As it turns out, SharePoint by default hides the duplicates -and doesn’t show you the ‘Show Duplicates’ link, which is hidden in the popup window for a search result anyway.
The effect was that while I was expecting 7000 results from a search, I only got 1 hit.

If you want to see all your results, you would ideally want to switch off some option on the results web part, to tell it to stop hiding duplicates. Unfortunately that option is not available on the web part UI (something Microsoft should really have put in there IMHO).

Luckily there are some workaround, and you can also enable the ‘Show Duplicates’ link, although I think that solution is far from ideal.
You can either change the TrimDuplicates setting in the web part code (by exporting the web part, changing the setting and re-uploading it), or change the query in the web part settings to make it GroupBy the property DocumentSignature.

Here are the relevant blog posts describing all this in more detail: