On April 16th, Microsoft launched Windows Azure VM into production and now offers new pricing options and larger compute instances.

There are now two super huge VM compute options (The A6 and A7) at 28GB and 56GB of RAM respectively! This makes it possible to run SharePoint 2013 in Azure (previously not practical during the beta due to the 14GB limitation) because SP 2013 runs better with at least 30GB of RAM.

The pre-pay and monthly contract commitment options are radically different than the flat price per VM model known as Pay-As-You-Go (which is still available). This is because of the introduction of compute quotas that I will explain below.

The pre-pay model is 40% less expensive than hosting VM’s at a colocation facility (which I will prove in this article) and it is also between 3x and 12x less expensive than the pay-as-you-go hourly model.

Let’s do the math…

The price per single GB of RAM per month is $38.28 in each model in the Pay-as-you-go, whereas it can be as low as $3.20 per GB if you go with the pre-paid 6 month contract because it includes a 20% discount.

image

http://www.windowsazure.com/en-us/pricing/details/virtual-machines/

Here is how the quota system works for the pre-pay and monthly contract commitment (6 month or 12 month) options work. Before any discounts are factored in, if you spend $500 dollars per month then you get a quota of 20 concurrent Standard Small (A1) compute instances, which means you can either have 20 small VM’s for $500/month, or you can mix and match and have two extra large “A7” instances with 56GB of RAM in each VM (that would use 16 of the 20 quota, leaving 4 that could be used for a ‘Large’ instance).

image

http://www.windowsazure.com/en-us/offers/commitment-plans#header-6 

To quantify the savings of this quota model compared to the hourly pay as you go model, I make RAM the common denominator for comparison. When you multiple the total amount of ram you get with 20 x A1 instances, you arrive at 35GB of RAM (20 x 1.75GB = 35GB total). Divide this by the cost and you end up with $14.28 per GB of RAM. That is almost three times less expensive than the pay-as-you-go price of $38.28 per GB of RAM!

It even gets better when you compare the amount of RAM you get with two A7 instances plus a Large instance: 119 GB of RAM for $500 per month, which equates to only $4 per GB of RAM/month! Add the 20% discount for a 6 month contract and you get down to $3.2 GB of RAM/month. So you get 3x more value with a few larger instances than multiple small instances.  And you get 12x the value compared to the pay-as-you-go plan.

You can tell Microsoft is taking this pretty seriously because the monthly and pre-paid plans are now up to 8x less expensive than the pay-as-you-go plans. This should attract a lot of customers, if they can understand the math and quota system. However, what is more exciting to me personally, is that the new plans are less expensive than hosting it yourself at a colocation facility (aka Private cloud).  Keep reading below for that comparison.

clip_image002

http://msdn.microsoft.com/en-us/library/windowsazure/dn197896.aspx 

The discounts increase from 20% to 32% depending on the amount of monthly commitment and pre-pay vs monthly.

image

How does this compare to on-premise hosting? If you host your servers at your own facility that you are already leasing for employee usage, then the average cost per GB is about $0.88 cents (excluding staffing costs) because you own the hardware and this assumes you depreciate over 3 years (to match most lifecycle replacement policies). However, if you host your equipment at an off-site colocation facility the price increases to $5.38 per GB /month. Therefore, it now costs $2.18 per GB RAM less to host on Azure than it does to host at a typical colocation facility (40% less!).  Public cloud hosting is now more affordable than owning the equipment and paying for a cabinet/power/cooling/internet at a colocation facility.

Now, none of this factors in the extra savings of eliminating the extra expense of backup, disaster recovery, security, bandwidth and Staffing, all of which make hosting on-premise much more expensive than public cloud hosting.

So is there anything preventing moving the whole VM infrastructure to Azure? Some might say that they still prefer Active Directory on-premise, but it is now possible to have a VPN tunnel to Azure Active Directory and create a VLAN for your Azure virtual machines. Here is a blog post from Sean McNeil describing those details. http://office365evangelist.com/?cat=26

So now what?

1. Integrate Active Directory with Azure http://office365evangelist.com/?cat=26
2. Deploy new VM’s in Azure
3. Move existing VM’s to Azure using System Center 2012 SP1 App Controller as described here:
http://www.techrepublic.com/blog/datacenter/deploy-an-on-premise-vm-to-windows-azure-with-app-controller/5919

You can also manually upload a VM into Azure as described by Adam Rafels here but it takes more effort, so I recommend App Controller.
https://www.catapultsystems.com/arafels/archive/2013/02/20/one-does-not-simply-upload-a-vm-to-azure.aspx