Calculating the variable cost of my Azure workload


rwagg-white small

Rob Waggoner

MS-Azure_rgb_Blk

Hello again,

Don’t worry, this is not a dive into the Azure calculator, but I am going to talk about planning the size of your Virtual Machines in Azure.  As you know, Azure has the ability to run your VMs just like you run them on-premises.  The big difference though, can come down to sizing.  Here’s my point.  When you plan hardware for an on-premises installation, you have to also think about the long term growth of the workload.  Typically, you plan for 3-5 years of growth.  The thing I really like about Azure is that you only need to plan for today!

For example, you build a new workload that will only be used by a few users at first and will initially only consume a little bit of disk space.  For an on-premises plan, you also have to ask, but how many users and how much storage will we need in 3-5 years?  And then buy for that capacity now.  In Azure, the story is different since you can add more compute or storage capacity on demand.  Now, we only have to build say an A1 Virtual Machine with a single Virtual Processor and 50 GB of disk capacity.  As the workload leaves limited rollout and becomes more widespread within the company, you may want to then increase the compute capacity to dual processing cores (an A2 VM) and add additional disk space to store the additional data.  With Azure, adding additional virtual processors just requires a quick reboot, and you can also quickly increase storage capacity.

Take this same scenario to a year from now.  Maybe you increase from an A2 VM to an A3 VM to provide four virtual processors, and your storage demand is now up to 750 GB.  The great thing about Azure is that you have this on-demand compute and storage capability, thus the customer only pays for what they use now, not what they need in the next 3-5 years. 

I talk about this on-demand capability often because it does change the way we size our workloads.

The other way to contain costs on your Azure consumption is to turn off your workloads when you are not using them.  We are all accustomed to running our infrastructure 24x7 even if we only use the infrastructure from 8-5.  In Azure, since you are charged for the resources consumed, if you turn off your VMs at say, 6:00 PM and turn them back on at 6:00 AM, you’ve now cut the cost of running that VM in half.  Then when you factor in that most companies don’t work on the weekends, you’ve now significantly reduced the cost of running your solution in Azure. Now, if you need to have your workloads running 24x7, do they need the same amount of compute capacity after hours?  Or are you just making sure that the random employee can have access at 2:00 AM?  You could also automate the size change of your VM around the workforce.  Maybe the VM is an A3 during the day, but you reduce it to an A2 at night to save some compute costs.

These are just a few of the ways you can leverage all the capabilities within Azure when you need them, but save on costs when the capacity isn’t needed.

 

Until next time,

 

Rob

 


Comments (1)

  1. Rusty Brown says:

    Rob, This is good insight and may solve my solution need. We have a few small business clients that the last items holding them back from being a Azure client is QuickBooks or MYOB apps. What is needed is an affordable hosting of Windows 10 or maybe an A2 VM running the app. We looked at RemoteApp but at a minimum of 20 licenses it isn’t for small business. Maybe you have another way??
    Rusty

Skip to main content