If we refer back to the ARM instance, we have a single ARM instance that manages, or essentially is that portal for the instance of Azure-Stack. Then you can have 1 or more regions below it. The same idea holds for Azure.

One or More Regions under the management of Azure Resource Manager (ARM)

So if you have a data center in US west, that shows up as a place where you can deploy resources. If you have a data center in US east, you can deploy resources there, and the same thing is going to occur with Azure-Stack. We will have those regions present under the ARM to be able to handle those as a single-entity of Azure-Stack (if you will).

Underneath each of those regions, there will be one or more Scale Units and we will go through and define what those are. You can consider them a group of computer storage resources and networking resources that interact together in a tightly bound fashion or delivered essentially as a unit or a whole and you are going to have one or multiple of those within a region to provide you to scale in the cloud. We are going to start off with 4 servers within each scale unit.

What does Azure Stack Region represent?

An Azure stack region represents essentially that group of scale units that share the same “physical location” – In the same physical location. You are going to have a set of racks of equipment if you have got some reasonable scale within a region and a set of scale units. They are probably going to be underneath a single administrator and the customer can decide what those characteristics mean. We have essentially the underlying requirement for a region; is that they have the network latency and bandwidth that match the anticipations of what you would see in an Azure Cloud Storage.

Networking Requirements

High-bandwidth and very low latency, because you are going to have workloads that are going to be essentially deployed within that region and have the requirements that they are going to be able to interact very effectively.

The network infrastructure needs to provide that bandwidth and latency. So you are not going to lay down a couple of one-gig-links between multiple scale units within a region, you are going to run ten gig, probably multi 10 gig configurations between those scale units to provide that.