Planning an Azure Stack Deployment
There are a number of items to consider when planning an Azure Stack implementation. Following are some points you should consider at a high level:
Scale: If your intent is to deploy Azure Stack for production workloads, consider the size of your deployment. This involves having multiple regions and multiple scale units per region if your enterprise workloads span several geographically distributed data centers in your organization.
Purpose: As discussed earlier in the “Designing an Azure Stack Scale Unit” section of this chapter, Azure Stack is an integrated system provided by Microsoft OEM partners, starting at a minimum of four servers in a scale unit. However, if you are using Azure Stack for evaluation and dev/test workloads, you can always deploy an Azure Stack Development Kit on a single box with reduced capacity and functionality. Microsoft will continue to ship preview builds for the Azure Stack Development Kit, so customers can test the new features before updating production Azure Stack systems.
Capacity: The IaaS instance offering in your Azure Stack deployment depends on the capacity of your integrated system. Factors such as the Azure VM sizes you want and the Azure Services you will offer and their requirements will determine the VM SKUs in production deployments. While future releases of Azure Stack may bring heavy VM SKUs to the hybrid platform, unless your data center can provide the underlying capacity needed, it would be useless to provision such SKUs without the required compute, storage, and network power.
Workloads: The workloads in your enterprise can be either IO-intensive, CPU-intensive, or memory-intensive. In Azure Stack, IO-intensive workloads require more flash devices as storage, CPU-intensive workloads demand higher bin CPUs in a scale unit, and memory-intensive workloads demand optimal VM density in each host.
Identity: Azure Stack leverages Azure Active Directory (AAD) or Active Directory Federation Services (ADFS) as its authentication mechanism. Multiple instances of AAD can be integrated to provide authentication for tenant services in Azure Stack. You also need to consider factors such as federated identity, self-service user management, and so on, to provide a robust end-user experience for your tenants.