Virtual Machine Availability and Downtime

Azure VMs can be affected by:

1. Unplanned Hardware Maintenance

   • Triggered when Azure predicts a hardware/platform failure.

   • Uses Live Migration to move VMs to healthy hardware, causing minimal downtime.

2. Unexpected Downtime

   • Caused by hardware/network failures.

   • Azure automatically migrates the VM to healthy hardware, but a reboot may occur, and temporary disk data can be lost.

3. Planned Maintenance

   • Routine updates to the underlying Azure platform.

   • Usually no impact on VMs.

   • Microsoft does not update your VM OS or software—this is the administrator’s responsibility.

2. Availability Sets

Purpose: Minimize impact of downtime by avoiding single points of failure.

• VMs in an availability set are distributed across:

  • Fault Domains (FDs): Physical separation (racks, power, networking). At least 2 FDs per set.

  • Update Domains (UDs): Logical groupings for rolling updates. By default 5 UDs, can configure up to 20.

Best Practices:

• Place multiple VMs in an Availability Set for redundancy.

• Separate application tiers into different Availability Sets.

• Combine with Load Balancer and Managed Disks.

SLA Guarantees:

• Two or more VMs across Availability Zones: 99.99% connectivity.

• Two or more VMs in an Availability Set: 99.95% connectivity.

• Single VM using Premium Storage: 99.9% connectivity.

3. Availability Zones

• High availability across physical datacenters in a region.

• Each zone has independent power, cooling, and networking.

• Minimum of 3 zones per region.

• Use zonal services (resources pinned to a zone) or zone-redundant services (automatic replication).

• SLA: 99.99% VM uptime when properly configured.

4. Scaling Concepts

Vertical Scaling (Scale Up/Down)

• Change the size of a VM (CPU, memory).

• Useful for under-utilized or resource-intensive VMs.

• Limited by hardware availability; usually requires VM restart.

Horizontal Scaling (Scale Out/In)

• Change the number of VM instances.

• More flexible in cloud environments.

• Supports autoscale to dynamically adjust based on workload.

5. Virtual Machine Scale Sets

Purpose: Manage a group of identical VMs with auto-scaling capability.

Benefits:

• Simplifies management of hundreds of VMs.

• Supports Azure Load Balancer (Layer 4) and Application Gateway (Layer 7).

• Autoscale adjusts VM count dynamically to meet demand.

• Supports up to 1,000 VMs (300 for custom images).

Configuration Considerations:

• Initial instance count, VM size, managed disks, Azure Spot instances.

• Spreading algorithm (max spreading recommended).

• Scaling beyond 100 instances may require multiple placement groups.

6. Autoscale

Features:

• Automatically adjusts VM count based on performance metrics.

• Scale Out: Increase VM instances when demand rises.

• Scale In: Decrease VM instances when demand falls.

• Supports scheduled scaling (fixed time scaling).

Configuration Parameters:

• Minimum/Maximum number of VMs

• Default number of VMs

• CPU thresholds for scale-out and scale-in

• Number of VMs to add/remove per scaling event

Benefits: Reduces cost and management overhead while ensuring application performance.