Microsoft’s Cloud Solution Provider (CSP) model has been in full force for about a year now, and at Convergence Networks, we’ve been delivering Office 365 solutions through the CSP program to deliver quick licensing and centralized management for our clients. The CSP program isn’t just for Office 365, though; it’s also available for many of Microsoft’s cloud offerings, including Azure. We recently had a fast-moving opportunity to deliver an Azure solution for a new client, so rather than going through traditional channels, we went through CSP.
Hitting the Road Running
A large not-for-profit organization approached us to do a build their new CRM infrastructure in Azure. There were two minor catches:
- The CRM vendor didn’t have any customers running their applications in Azure, so couldn’t give us any guidance on how to provision for anything other than an on-site infrastructure (physical or virtualized).
- The entire infrastructure had to be designed, implemented, and ready to hand off to the CRM vendor in 10 days.
Entire infrastructure project in Azure in Ten. Days.
“No problem!” you say. “After all, that’s what public cloud is about. You just log in and start spinning up VMs in minutes.”
Of course I said, “Sure thing!” I like a challenge.
Design Phase: Unknown Unknowns
Day 1: We got a Visio diagram and a deadline.
The diagram depicted what the production infrastructure would look like if it were deployed on-site — servers, cores, RAM, disk capacity, and IOPS for the OLTP and OLAP servers. There were a couple of load balanced web servers and load balanced app servers in the mix, as well.
The specs in the diagram called for everything needed to run SQL Server Enterprise — OLTP, OLAP, and SQL Server Reporting Services (SSRS) are key components of the entire CRM.
Network diagram in hand, I headed over to the Azure Calculator that Microsoft kindly provides.
A few hours of research and filling in numbers later, and I had a rough estimate of the monthly cost to run the infrastructure full time: every machine powered on 24×7 with the vendor-specified RAM, CPU and storage, along with networking, load balancers, static IPs, and even estimated storage transactions.
Estimated cost: About $18k/month.
Lesson 1: “It’d Be Cheaper…”
I ran through the Azure calculator several times before talking to anyone about the estimated cost. Surely I was wrong. But no: the big VMs, the ones that need 192GB of RAM — those are expensive when loaded with SQL Server Enterprise licensing. Two VMs accounted for about $8k each of cost.
Across the board, once I mentioned this internally, the response was along the lines of, “It’d just be cheaper for them to buy the infrastructure outright.”
And while that might be true – though SQL Server Enterprise is far from cheap for the number of cores the project called for — it’s part of the first lesson. Cost isn’t the sole factor for people making the leap to the cloud. If cheaper was always the right option, Convergence wouldn’t be in business.
So, as jaw-dropping as that figure might be at first glance, the client had already made the determination that Azure was their destination before they’d engaged us. They’d looked at the Azure calculator, too, and had arrived at the same rough cost estimate, and decided that the value that Azure provided – the ability to run their entire infrastructure in the cloud without having to purchase hardware, software, installation, and ongoing maintenance – was worth the cost.
I did make a call to the client just to check and see if that was in line with what they expected, because if not, we could call it quits here.