Windows 365 service resilience
Windows 365 is designed to provide a resilient and reliable service for organizations and end users, connecting to, and using their Cloud PCs.
Windows 365 uses the Azure Virtual Desktop service to connect and broker end user connections to their Cloud PC in any of the supported Azure regions from anywhere, on any device. To minimize outages and support end user and administrator requests, resilience is architected into these services. In addition to this, Windows 365 operates another set of platform infrastructure that provides the many administrator and end user features that manage and control the overall Windows 365 experience. Microsoft fully manages this infrastructure.
Windows 365 is a software-as-a-service (SaaS) service, where customers of the service don't have to architect, deploy, or manage complex infrastructure. As a SaaS service, Microsoft manages the underlying infrastructure for all of the individual Windows 365 services. This management maintains a global, fully available, resilient service that customers can rely upon for the daily use of their Cloud PCs. Microsoft continually works to improve the architecture of the service to improve the resilience and recovery times should there be an outage in any of the related Azure services. However, as a SaaS service, Windows 365 requires shared responsibility in overall service delivery to end users. Windows 365 aligns to the Microsoft 365 guidance and recommendations for business continuity and cloud partner responsibilities
In addition to the Azure Virtual Desktop connectivity layer, Windows 365 operates many dedicated services that are architected as microservices to support agile operations and independence. Each of these services performs specific tasks related to administrator or end user requests. Some examples include:
- The Windows 365 provisioning service processes deployment requests to provision Cloud PCs according to the specifications and locations defined by the administrator making the request.
- The Entra ID notification service monitors administrator changes to Microsoft Entra ID groups, subscriptions, and licenses that trigger new deployments.
- Other services manage:
- The administrator portal.
- The end user portal.
- Disaster recovery.
- Diagnostics.
- Troubleshooting.
- Service monitoring.
- And all other functions of Windows 365.
Each of these services:
- Is its own web service that uses standard Azure infrastructure services. Each service has a certain set of extra Azure infrastructure requirements like CosmosDB, Azure storage, and Event hubs.
- Is architected for resiliency. For example, for the services storage requirements an Azure Storage account with Globally Redundant Storage (GRS) is used. For database services such as CosmosDB, the data store is replicated across regions.
- Is architected to use Azure resilience services like Azure availability zones, and cross region failover.
The following diagram shows the architecture of an example service. Windows 365 distributes its infrastructure across multiple availability zones within a region and across multiple Azure regions. The service operates in an active-active manner. This supports in-region and cross-region resiliency. If an outage occurs within a region, the service continues functioning. If a region fails, the service is transferred to the secondary region's infrastructure, and normal operations continue.
Example architecture of a Windows 365 service that enables administrators or end user operations:
Virtual machine resiliency
Each Windows 365 Cloud PC is a single instance Azure virtual machine. Resilience is provided at the Azure host level to mitigate any compute continuity issues. For more information, see Business continuity and disaster recovery overview.
Client resiliency
Windows App is the Windows 365 client used to connect to your Cloud PC. Windows App can also be used for connectivity to any Windows in the Cloud service, such as Azure Virtual Desktop, Microsoft DevBox and Remote Desktop Services, as well as RemotePC or direct RDP connections. Windows App is regularly updated to provide new functionality to end users.
There are two deployment mechanisms to test the client after it has gone through extensive internal testing:
- An Insiders ring where the client is tested by a set of users who have opted into this ring. Feedback is then provided and bugs can be fixed before rolling out to the Public ring.
- Windows App is hosted within the Windows Store, which enables a reliable global software distribution service. The Windows App uses the Microsoft Store gradual package rollout which lets Microsoft update groups of machines over a specified time period. By monitoring the feedback and usage analytics of these initial groups, we can ensure that the update is ready before rolling it out to a wider community. For more information, see Gradual package rollout.
Customer best practices
As Windows 365 has a shared responsibility for overall service delivery, there are some best practices that customers should follow to increase the resiliency of the overall service.
Microsoft hosted network. Use a Microsoft hosted network when creating a provisioning policy. Microsoft hosted network is a Microsoft managed networking option, where no Azure vNet or subscription is required for Cloud PC network connectivity. This option lets Microsoft make the placement decisions for the Cloud PC virtual machines and decreases the possibility of provisioning issues. For more information, see Microsoft hosted network
Azure Network Connection. If you need granular control of your corporate network traffic – firewall rules, custom routes, or on-premises network access – the Azure Network Connection (ANC) feature lets you bring your own Azure vNet/subscription to Windows 365. For customers using the ANC, create multiple connections and prioritize these connections in order of failover usage. This prioritization provides a network connection failover if there's a networking outage. For more information, see Azure Network Connection
Point-in-time restore. This service allows you to fully restore your Cloud PC to a previous state. You can configure the service to automatically create restore points across short and longer time windows. For more information, see Point-in-time restore
Cross region disaster recovery. This Windows 365 add-on feature creates snapshots of Cloud PCs. These snapshots are placed in customer-defined, geographically distant locations. The snapshots can be recovered to Cloud PCs running in the selected location during a disaster recovery event. For more information, see Cross region disaster recovery