Last month, Microsoft introduced a number of security enhancements for Azure, and also unveiled reserved capacity offerings for six more Azure services. Today, the tech giant has announced the preview of Spot Virtual Machines for its cloud computing platform as a form of replacement for low-priority Virtual Machines on scale sets (VMSS). Essentially, these enable Azure users to deploy workloads across unused parts of the offered capacities at deeply discounted prices.
Spot Virtual Machines will allow for usage of the same features that are made available through pay-as-you-go VMs, at lower prices and varying eviction conditions. Workloads that have noted as suitable for these include batch jobs, short-lived jobs with easy operational recoverability, development and test, and more. Notably, Spot pricing will be available both for single VMs, and for VMSS.
As far as the pricing is concerned, it will be less volatile in comparison to that of low-priority VMs to support better budget-management. Continuing along the same route of a different approach, pricing for Spot Virtual Machines will also vary based on capacity for different sizes and SKUs. All varying prices based on region and sizes will be displayed on the Azure Portal in order to help customers make a more informed decision with regards to their choice of VMs.
To make things easier for customers who use different types of VMs, the Spot Virtual Machine quota will be different from the pay-as-you-go VM quota. With regards to the handling of evictions, Microsoft has warned that runtime is not being guaranteed, so customers should be well-prepared for evictions at any point when certain conditions are not being met. However, subscribing to scheduled events can get users a 30-second notice before eviction. The aforementioned conditions include Spot prices going above the maximum defined price, and Azure needing to reclaim capacity.
You can read in detail about the deployment techniques and best practices recommended for better management of Spot Virtual Machines at the original blog post.