You can read the official announce Announcing general availability of Managed Disks and larger Scale Sets
I had the opportunity to test them during the devlopment process and I must say they are a giant leap forward in terms of storage manageability.
Managed disks are a first-class “disk” object for VMs. They are fully managed, meaning storage availability, access security, and scaling are handled for you by the platform. A managed disk attaches to a VM, whether the OS or data disk, and will not be exposed as a blob in your storage account but instead as an “disk object.” You can manipulate it like other ARM objects via API/CLI/PS and portal (including RBAC). Here are some of the specific characteristics:
- Managed Disk storage is fully managed for you. You no longer have to worry about over-filling your storage account with too many VHDs and accidently throttling your VMs. You can create as many Managed Disks as you want (up to 5,000 per region) without worrying about creating more storage accounts (or running out of storage accounts in a subscription).
- Managed Disk availability is also fully managed. Managed disks automatically spread the disks attached to VMs in an availability set across distinct storage stamps/deployments. Thus, any storage failure will only impact one VM in an availability set as we will make sure the “fault domains” of the VMs include both the compute node and the storage accounts.
- The disks that are attached to your VMs do not have a publicly accessible endpoint. They are now ARM-based resources (which means they have RBAC on their APIs) instead of a SAS-based storage REST endpoint with internet access.
- Given the scalability of these disks, they also simplify VM Scale Sets scalability. With the launch of Managed Disks, Azure will support attaching data disks to VM Scale Sets (a common request) and will support up to 1,000 instances in a VM Scale Set.
- Azure will support migration from existing storage VHDs into managed disks, to make it really easy to move to the new model.