The end of the discrete Virtual Machine pricing

Or, the impact of the Ignite’21 release of the Azure Virtual Machine Scale Sets flexible orchestration on pricing levels.

Less than 20 years ago, each application ran on its own physical machine, with dedicated physical processors, internal memory and a hard disk, where the price was determined by the type and number of processors, the size of the internal memory and that of the hard disk.

Nothing has changed after the introduction of Virtual Machine in the pricing except for the price level and an Azure VM is still sold as that physical machine from 30 years ago. Pricing maintains the illusion that an Azure VM is a discrete component based on the completely misguided assumption that a 365/24/7 computer is loaded in the same way.

The Ignite’21 release of the Azure Virtual machine scale sets with flexible orchestration is very interesting from a cost reduction perspective and heralds the end of discrete Virtual Machine pricing. With orchestrated scale sets, a virtual machine becomes truly virtual and can change in resource size and priority at any time based on metrics, schedule or AI predictions.

This means you only pay for the necessary instantaneous size of your virtual machine. This can lead to significant price reductions when you realize that the “365/24/7” machines are generally heavily oversized, up to 80%.

The big question then is, how do you calculate the effect of Azure VM scale sets on pricing? It is clear that the current way of pricing Azure VMs as discrete pricing items is no longer adequate. The settlement of consumed capacity of certain VM series is then a better fit.

Using the fact that the price structure of a VM series is linear and the GBRAM/vCore ratio within a series is constant. With this you can, for example, see the amount of ‘used’ GBRAMs as a unit price, as for example the electricity power consumption is expressed in kWh.

Quotations thus become an pricing estimate of the predicated ‘consumption’ of GBRAMs. For a further explanation of this concept, see the video I made about this topic: