Last week Microsoft released the public technical preview of new Azure Stack. Azure Stack, along with its predecessor Windows Azure Pack, gives anyone the ability to extend Azure management capabilities to their on-premises datacentre.

Firstly, a bit of background.

With Windows Server 2012 R2, Microsoft made available Windows Azure Pack. Azure Pack offered an on-premise integration point between Windows Server, System Centre, and SQL Server to offer a self-service portal and private cloud services including virtual machine provisioning and management (IaaS), database as a services (DBaaS), and scalable web application hosting (PaaS). Azure Pack can be integrated with Service Management Automation (SMA) and leverages Desired State Configuration (DSC) to manage IaaS deployments.

Microsoft built Windows Azure Pack using a common Service Management API based on the Azure Service Management (ASM) API which is how the “Classic” Azure platform is managed. Azure Pack delivers a Self Service portal based on System Centre Service Provider Foundation (a subset of Orchestrator features) that utilises templates and other library resources from System Centre to provision virtual networks, virtual machines, web apps and other IaaS and PaaS resources on-premises.

It’s about the Stack

The rapid rate of change in Azure and the introduction of a new management API (Azure Resource Manager – ARM) meant Azure Pack features fell behind what is possible in Azure today.

In 2015 Microsoft announced that with the release of Windows Server 2016, they will provide Azure Stack, which is the next iteration of hybrid cloud integration, looking to address Azure Pack’s shortcomings by providing an Azure-consistent API (using ARM) for managing underlying resource providers (compute, network and storage) across public and private clouds.

By utilising the ARM API across both Azure cloud and Hyper-V on premise, Azure Stack users will be able to:

  • Deploy resources across public and private clouds from the same JSON templates and PowerShell Cmdlets
  • Reduce reliance on System Centre to manage underlying resources (ARM does not use System Centre on-premises, though System Centre may still be used to manage infrastructure)
  • Utilise the same resource management model over public and private clouds.

Adopting a Hybrid Cloud approach marks a movement away from integrating hardware with software and towards making software control the underlying hardware, without regard to it, other than as additional capacity.

Now, when we consider the design of an on-premises Private Cloud, we know it needs to take into account the Azure tenancy and the toolsets and processes that have been implemented or are planned for the near future. Having this inter-operability between platforms empowers us to extend a true ‘Hybrid Cloud’ for greater cloud consistency.

It is important to note that at launch only a fraction of the full Azure offerings are available on the Azure Stack preview. At this time this includes compute, storage and network resources (IaaS) plus some of the PaaS features such as web and mobile app services.

Aside from feature gaps, there are also markedly different use cases for deploying applications to on-premises or to the cloud, such as

  • Compliance related concerns as to where your customer or corporate data resides. For some industry sectors it can sometimes be a lengthy process to get cloud use approved by legal, regulatory, security and industry stakeholders.
  • Financial constraints can determine one approach or the other – if deploying on-premises infrastructure, you will have to deploy for peak demand and if this is only intermittent (e.g. monthly or quarterly usage spikes) this can provide a better case for scalable cloud services. On the other hand, for static loads that need to run 24/7, on-premises can be a more cost effective approach (although it can also be misleading to compare without including sunk costs of the datacentre vs. the prospective cost of a cloud service… but that’s an argument for another time).
  • The other big reason for deploying workloads on-premises is that your organisation’s journey to the cloud is going to be a long one, so while the aspiration is there, a pragmatic approach is to deploy the easy workloads to the cloud first and the more difficult workloads and applications remain on-premises.

As most large organisations face one or more of the scenarios above, Azure Stack delivers the benefits of cloud manageability, speed of deployment, and the economy and simplicity of having one management framework across the Public and Private clouds.

One the great use cases that springs to mind is to deploy your dev/test workloads to Azure and take advantage of scaling up during heavy use and UAT, but when there is little activity simply turn off your servers and automate out of hours and weekend shutdowns to further reduce costs. When code is ready to be deployed to production you can then use all the same templates to deploy into your production systems that reside in your datacentre hosted on Azure Stack.

To get started, download and deploy Windows Server 2016 Technical Preview 4:

Azure Stack can be downloaded by registering at the following link:

And you will have to be mindful of the Azure Stack requirements:

That’s it for now, but there will be much more to come on Azure Stack as we explore the platform and start rolling out with our customers.

Architecture, Azure Infrastructure, Azure Platform
, , , ,