A little more over a year ago Microsoft started shipping Azure Stack. The ability of running a subset of Azure features running in your datacenter was (and still is) just amazing. But what lies ahead of us in the near future? Slowly the role fit for Azure Stack in the Azure eco system settles in and its one I didn’t anticipate over a year ago!
Azure: The World’s Computer
What is Azure World’s Computer? Where is this coming from. A nice quote (or wrap up ) was done in this article and stated:
“Nadella’s “cloud as the world’s computer” begins with Azure metaphorically as the OS, continued with cloud-based products and services as the programs and the intelligent edge and diversity of edge and IoT devices as the UI.”
The quote from Nadella sums it all up, but a view of the distributed layers will always help. See below that vision in a picture:
On the left top we see Azure Sphere, the Linux distro build by Microsoft to run on microcontroller units (MCU’s) and is being managed by Azure. This MCU can run your coffee machine by the time you are reading this. Drones that use ML techniques that run in a container in the drone to recognize a particular situation and take action upon that. Same for industrial assets that need close to the machine/device computations to take action for the machine based on an algorithm. Azure Data Box Edge is a storage solution that allows you to process data and send it over network to Azure and during transfer perform ML based algorithm over that data.
And then there is Azure Stack. Azure Stack allows you to have a subset of Azure services running in your own datacenter. As with all the other edge devices, meant to transfer data to Azure eventually, I believe Azure Stack will do the same. Its bigger than a databox and definitely has more computing power. So, imagine having Tesla or any other autonomous driving car having data about the environment and there is 5G that can connect these cars to a farm of distributed services running on Azure Stacks that are deployed at 5G antennas in the country. Services running on Azure Stack will allow communications to vehicles and, because of the intense compute power available, generalize and upload the data to Azure Storage for big data analysis. The ‘service’ the car connects to will have redundant endpoints to provide high availability.
Then all the Azure Services to manage it. Security Center & AAD for security and identity across edge and cloud. A unified management experience and the Partner Eco system for reselling / co-selling. So, that’s the horizon. The point we are getting to along the days, months and years.
Azure Stack and scale
When Azure Stack was announced at Ignite back in May 2015 my architect mind started spinning creating awesome look alike Azure Topologies. It was about having a single portal endpoint like portal.azure.com, but then for my portal.azurestack.nl. The next phase was about building resiliency like that was done in Azure with regions like West and North Europe. The notion of scale units came across at a later stage to provide multiple clusters in a single region… Then Ignite 2018 appeared and the story above about Azure: The World’s Computer sank into my brains… You most likely have these aha moments as well, but for me I concluded that there is a new path for Azure Stack and a repurposing of the solution. In 2017 scale was on the roadmap for FY18H1. It didn’t come… And it still is not there today. Its planned, for sure, but to my honest opinion, multiple scale units (hopefully) this year but what about multi region? I think multi region doesn’t fit in the Azure World Computer story. As explained earlier, the Azure World computer must be seen as a whole. Azure Stack is in the cloud native application landscape a broker to connect the edge back to public Azure, whereas, with some level of redundancy, a subset might reside on Azure Stack for the lifecycle of the application. Even in some cases a fully disconnected Azure Stack is needed to fulfill specific requirements.
Focus for Azure Stack
One mans’ opinion here… In the Azure World Computer, computation also needs to be done at the edge. Still, a consistent platform is key to success in order for application developers creating apps that run native on Azure or all the way at the edge. With the rise of 5G, I think Azure Stacks’ edge computation power becomes more and more important for example like autonomous driving and for smart cities. In order to work with data at the edge, native Kubernetes service support like we have in Azure today is a really important feature. In order to run some analysis at the edge, I also believe Azure Data bricks needs an entry point in Azure Stack. Next to the features Event hub and IoT hub that are currently in private preview, I think the world computer starts serving its purpose once those features are implemented and general available (GA).
The pitfall for Azure Stack
Be good at what’s needed for now. No wait, be excellent for what’s needed today. Let’s face facts, how many companies/ enterprises are ready for this new world computer? A lot of them still runs virtual machines -Simple fact-. And those fat big monolithic apps need resiliency in the infrastructure layer. Stuff that IT Pros have done for years in virtualization environments. Azure is providing really good HA/DR solutions to serve the current customers and enterprises, but this needs to ‘up level’ in Azure Stack. AWS announced AWS Outpost and remember back somewhere in 2008, where Microsoft told the world they needed PaaS? It turned out we wanted ‘stateful’ virtual machines. AWS offered this and they won the first battle. Today a lot of customers still want this and also in Azure Stack. Microsoft shouldn’t lose focus on the current ask. Otherwise AWS will do and (might) gain enterprise market. An advantage of Azure Stack is that it can run fully disconnected and AWS Outpost doesn’t. On the other (Cloud native) hand, to gain the advantage we do need more and more ‘as a service’ offering in Azure Stack too. So my priority would be on both of these areas; Enterprises that want to operate a cloud themselves and transform on their own pace and first class born in the cloud apps that needs to run from public cloud all the way to the edge.
Update 19-03-26: Azure Stack HCI is the rebrand of WSSD and adds a new perspective on the pitfalls mentioned above. Read more here
every year 9 billion new devices are shipped. These are sensors, controllers, monitors each with their own processors. You find them in cars, factory floors, HVAC systems, and consumer appliances… the sum of this processing power at the edge is greater than all the public clouds.
Satya Nadella – Ignite 2018