Licensing costs will skyrocket to indefensible levels under new method of charging for memory usage.
A user at a large public university said: "By my calculations, which have been confirmed to be accurate by VMware, upon upgrading to vSphere 5 we will have to purchase 19 additional VMware licenses to make use of the memory in our systems. This is a purchase to maintain access to the memory that we are allowed to freely use in vSphere 4.x on the same servers. Based on pricing data from our last license purchase, this will be a cost in excess of $53,000.00."
That user goes on to say: "No matter how VMware spins the change, it boils down to the fact that they have placed an arbitrary limit of 48 GB of memory per CPU that did not exist in the previous version. It is very unfortunate that they have chosen to do this at a time when offerings from their main competitors, Citrix and Microsoft, are coming closer to matching VMware vSphere in functionality. While we do not currently have any plans to change primary virtualization vendors, this does make us more likely to entertain expanding the presence of Citrix XenServer in our data center."
As I see it, there's nothing wrong with VMware's concept of charging more as users use more resources. The problem is that the vRAM allocation per license is ridiculously low and will end up being a dollar for gigabyte Moore's law penalty to users as they take advantage more capable hardware. With vRAM allocations so low per license, the per CPU costs essentially vanishes, or will soon with Moore's law, the pricing model is based solely on total memory in use. Double your memory, and you'll double your licensing costs. That sort of constraint will surely be enough to drive at least some users to consider competitors with more reasonable licensing policy.
It seems to me that total RAM used isn't necessarily a bad basis for licensing a product like VMware, but the linear pricing model is unfair. Using the Enterprise Plus pricing, a more appropriate calculation might be something like the current $3,495 per socket with a 64-GB vRAM entitlement, but then sell additional vRAM entitlements for something like $500 for another 64 GB.
In Tuesday's announcement, VMware bragged about increasing its virtual machine capabilities so that users could configure very large VMs able to take on most any workload. The new maximum limits per VM are 32 processors and 1 TB of memory. What the company didn't mention is that under its new licensing, customers will need 22 Enterprise Plus licenses at a cost of $76,890 just to run one such virtual machine, whereas under vSphere 4.1, the largest virtual machine possible, with its 8 cores and 256 GB of allocated memory would have cost $3,495. Under the licensing model I suggest above, the price would be $10,995. You decide which is fairer.
While VMware was hardly forthcoming with the ramifications of its pricing model in Tuesday's presentation, there's plenty of information and examples of the new model on its website. As users examine the model and assess their own installations, most seem to feel they'll at least need to double their VMware licenses, and that that'll be compounded over time as Moore's Law bumps up standard RAM capacities.
And that's the problem with a policy like this. It seems to be based on usage patterns that were common two years ago. Many users now configure servers with far more RAM than VMware thinks they do, and in another 12 months, most every user will be affected by the policy as standard server RAM configurations continue to climb.
I spoke to Timothy Stephan, senior director for product marketing at VMware, and with Alberto Farronato, also a product marketing manager at VMware. They confirmed that with its new licensing VMware is intending to move to charging based on usage, in this case based on total vRAM allocated across all licenses owned by an organization. Stephan went on to say that VMware feels the model more accurately reflects the sort of usage-based charges that are usually thought of in cloud environments. He says that with the variety of workloads, and the use of such techniques as disaster recovery sites and high availability clusters, most customers shouldn't see substantially higher charges for licensing, and that often-times vRAM totals will not force users to purchase more licenses.
When I asked about the notion of a Moore's Law penalty based on every increasing RAM densities, Stephan and Farronato seemed to accept that effect, and felt that as hardware configurations increase in resources, VMware should also be paid more. I seriously doubt any of his customers are going to agree.
If Microsoft and Citrix were looking for a way to entice customers away from VMware, they were just handed one--in a Tiffany box, on a silver platter, all wrapped up and ready to serve.
Art Wittmann is director of InformationWeek Analytics, a portfolio of decision-support tools and analyst reports. You can write to him at firstname.lastname@example.org.
To find out more about Art Wittmann, please visit his page.
Data centers face increased resource demands and flat budgets. In this report, we show you steps you can take today to squeeze more from what you have, and also provide guidance on building a next-generation data center. Download it now.
InformationWeek Must Reads Oct. 21, 2014InformationWeek's new Must Reads is a compendium of our best recent coverage of digital strategy. Learn why you should learn to embrace DevOps, how to avoid roadblocks for digital projects, what the five steps to API management are, and more.