Up until now, Microsoft has guaranteed a 99.5% uptime SLA for IaaS (Infrastructure as a Service) virtual machines. While this level of uptime is very good, it is not enough for critical applications which require 24×7 access. An online retail app on Black Friday dealing with any downtime can mean big losses in revenue.
Why was there a 0.5% downtime expectation? When it comes down to it, the Azure servers which are hosting virtual machines are running Windows (of some sort). As such, they occasionally need to be patched with security and vulnerability updates. Just like any Windows machine, Windows updates often require a reboot. When these Azure hosts required a reboot, all of the virtual machines they were hosting would become unavailable.
Previously, in order to achieve a guaranteed uptime SLA higher than 99.5%, you would need to run multiple virtual machines in an availability group. For small businesses and small installations, this effectively doubles your costs and complexity.
This week, Microsoft’s Cory Sanders announced “a new 99.9% single-instance availability SLA to better support applications that cannot easily scale beyond single VMs”. There is a prerequisite however: All storage disks should be using the faster and higher priced premium storage. Premium storage allows up to 80,000 IOPS and throughput speeds up to 2,000 MBps.
This is just one more example of how the Azure cloud is ever-evolving, and getting better all the time.