Is it worth it? Azure Reservations and Savings Plans VS PAYG
In my work helping Azure customers optimize their workloads, I often get this question: how do I know when it's worth reserving or shutting down my virtual machines?
Well, thankfully, math can help us here with a simple calculation for the break-even hours.
But first, I must say that the cheapest option should always be always considered, and it's very often the "on-demand" use: switching on resources only when they are needed and off once done. There are several benefits to this policy, including instilling in the cloud teams the notion that they should only allocate resources when they are used and needed, and not just keeping them idle "just in case".
Let's do the math. I will pick a VM size, for example a E32ads v5. The current price listed for East US is as follows (Pricing - Linux Virtual Machines | Microsoft Azure ):
The monthly rates from the same page are:
so the discount is very good, especially with 3 Years Reservations! But I still want to know if it may be more convenient for me to switch it off and on when needed, thus not using it 7x24 each month.
The calculation is:
领英推荐
"hourly cost Reserved" *24/"hourly cost PAYG"
so in this case = $0.7965*24/$2.0960 = 9,12 hours is the break even
so the cost for 9,12 hours in PAYG is $2.0960*9,12*30 = 573,4656$ for the month, which is less than the Reserved monthly price.
This means that to make it worth reserving I need to be using that VM more than 9,12 hours a day: using it for 9,12 hours or less , PAYG is cheaper.
Let's do the calculation for Savings Plans now:
$0.9799*24/$2.0960 = 11,22 hours
Since the discount is lower, I have to make sure my VM is used more than 11,22 hours a day to make the savings plan cost effective. In fact, the cost for running the VM for 11,22 hours a day would be 705,5136$, less than the Savings Plans' monthly rate.
My key takeaway is: please always consider on/off policies where possible, as these have an impact not just on cost but also on the used resources and ultimately on their carbon emissions.