Stateful Node for Azure
When running stateful applications in the cloud, it is more likely you’ll find them using pay-as-you-go (i.e., on-demand) VMs, which will stay up and running throughout the application’s lifetime. On the other hand, using pay-as-you-go VMs typically cost a lot more than spot VMs. As these applications grow, so too does the cloud bill, and finding ways to optimize costs is becoming a strategic goal. With the potential to save up to 90% on cloud compute, it is hard to overlook the benefits of Spot VMs for these workloads.
With Spot, you don’t have to choose between resiliency, performance, or cost. You can easily run stateful workloads using spot capacity. With Stateful Nodes, you can have cost-efficient, reliable infrastructure that also keeps storage and IPs persistent at all times. You can utilize Azure’s spot VM technology to save up to 90% of your infrastructure costs while running stateful applications without any concerns.
Common Use Cases
Managing your stateful workload using a single instance solution is a very popular approach for many compute workloads such as:
- Personal developer machines
- Single DB instances
- Machine learning environments
- Applications that require stateful (IP and storage) capabilities
Stateful Persistence Options
The Stateful Node feature provides the ability to persist your OS disk, data disks, and network.
Persist Data and OS Disks
Data and OS disks are maintained using one of the following methods:
- Re-attach: The same disks are detached from the original VM and reattached to the new VM. If the new VM is launched in a different availability zone, a new disk is created from the latest snapshot and attached to the new VM. (Disks cannot be migrated between availability zones.)
- Take Snapshot: Periodic snapshots of the data disk are taken while the VM is running. Upon VM replacement, a new disk is created from the latest snapshot and is attached to the new VM.
Persist Network
New VMs are provisioned with the same private IP and public IP using the same network interface.
