Yesterday Microsoft published an update to the SLA for virtual machines running on Azure. As you probably know, Azure didn’t provide an SLA for single instance virtual machines: the 99.95% uptime guarantee was only available for at VMs that are part of a availability set or scale set. I blogged about this in a previous article.
With the latest update, Microsoft now provides an SLA for single instance VMs:
- An SLA of at least 99,9% is offered for single instance virtual machines;
- To qualify for this SLA, all disks of the single instance virtual machine must be using premium storage (providing 80.000 iops and 2.000 Mbps throughput)
- Microsoft can announce “Single Instance Maintenance” at least 5 days in advance. This downtime is related to hardware, network or service maintenance and will impact the availablity of your single instance VM.
So, this is an interesting improvement. However, you have to evaluate if 99,9% availability is a good enough SLA for your VMs. This imposes around 45 minutes of downtime per month. If you need a higher SLA, you have to work with availability or scale sets. Also take in account the premium storage requirement and how you will deal with Single Instance Maintenance.
Notice that Microsoft didn’t implement some kind of live migration technology into azure to achieve this SLA, Microsoft is using machine learning technology to predict hardware failures. The premium storage requirement also has to do with the high availability of this type of storage.
I hope this was helpful! The full SLA is available here.