The news is out, VMware has changed its vRAM licensing model for vSphere 5. VMware has clearly decided to continue its premium pricing strategy for what is hands down the best virtualization platform on the market. VMware probably views this pricing as necessary to continue to fund the massive investment in R&D that is required to maintain this leadership position, and to protect itself from the ever increasing levels of workload density that are being enabled by ever more powerful CPU’s and ever greater core density.
Alan Renouf has updated his License Validator script to reflect the new model, so perhaps it is time to run some new numbers. But first some bits of history:
- ESX v2.x to Virtual Infrastructure 3 (VI3) upgrade went from 2 sockets for ~$X to 1 socket for ~$X, which was a doubling in price and introduced different license levels.
- VI3 went to vSphere 4 where VMware introduced two more license levels Advanced and Enterprise Plus. There was a commensurate cost increase for those products over Standard and Enterprise. There was also an uproar over VMware’s attempts to drop the Enterprise license level, they reneged and Enterprise stuck around. This still impacted every one of their current Enterprise License Agreements (ELA) for those who wanted the latest features of Enterprise Plus. The late license change protected those who already had an ELA for the Enterprise license level.
- vSphere 4 upgrade to vSphere 5, VMware introduced vRAM Pool licensing, which conservatively would only double the price of vSphere for some, and quadrupled the price for others. Now there is another late license change that protects those with currently maxed out 2 socket half-height blades (max memory of 192G) and Enterprise Plus licenses. For those with Enterprise Licenses, there is still a possible price increase.
So historically, we see that VMware has been increasing the price of its software at every major release. Up until these vRAM changes, the changes in VMware’s pricing balanced out progress in server hardware, and for the most part did not create huge increases in the cost of virtualizing workloads for customers (per host licensing costs have been going up, but advances in CPU power and core density meant that fewer hosts have been needed over time).
vRAM pricing broke the historical pattern because it allowed VMware to charge for something that was not getting twice as dense and efficient every 18 months (CPU price/performance due to Moore’s Law). Even with these latest changes some important questions remain. Will the migration to vSphere 5 be a 1 to 1 upgrade with no incremental licenses needed for your environment? If not, are the extra features and capabilities in vSphere 5 worth the extra cost? If they are worth the extra cost, is this true across your entire environment or only in certain places? What is the cost of going to a multi-vendor strategy for virtualization platforms (and having to then manage that) compared to the cost of paying more for vSphere 5?
So the first question to answer is whether or not for your environment (and licensing arrangement with VMware) is this a simple 1 to 1 upgrade with the addition of Service and Support or something else? If it is something else, this could adversely impact VMware in the following manner.
VMware will have a loss in projected revenue as people may not upgrade immediately or at all. Here is an example:
On a simple 2 node cluster required by security policy and politics using its own vCenter for separation of duties ~88G of memory is configured currently.
- In the new model, assuming 44G is configured for each node, the lower end licenses would not cost an extra license per node just to perform an upgrade as there is now 32G per socket (64G per node)
- In the new model, assuming 44G is configured for each node, the Enterprise or Enterprise Plus licenses would not cost to upgrade.
- The hardware upgrade path for these nodes was to add more memory to grow them to 192G of memory over time while adding more VMs as at the moment their CPU capacity is around 10%. So there is plenty of room for VM growth on these nodes. This upgrade approach works fine for Enterprise Plus and may work for Enterprise but could require additional licenses of lower end licenses to also do this upgrade.
So if nothing changes in these boxes but memory, VMware will want more money? Why is this? Is VMware thinking that they are a service provider who can charge based on utilization they may or may not be able to measure due to 12 month highs and lows. They would charge based on the high-water mark average, so only looking at the peaks and not the valleys of configured memory. Does this also mean, that in order to use vSphere 5, Enterprise environments will need to also use VMware’s vRAM memory chargeback appliance used by the service providers?
The new changes help those with Enterprise and Enterprise+ but do not have a dense environment. It still impacts those who use the lower-end licenses but have configured more than 32G per socket in their clusters (except for ESXi free which is 32G per node). It still ignores the political necessity of separate environments with separate vCenter’s within an organization so that pooling of resources does not help.
Bottom line, there is still an impact on VMware and people are still looking at other alternatives for their hypervisor and cloud stacks. Here are the basic numbers for a small environment that scales up as you add more multiples of vRAM levels, CPUs, and servers. In this example, we are using list prices and fitting our workloads into the available vRAM, as well as the necessary management tools to manage the environment.
|Microsoft Hyper-V||Citrix XenServer||VMware vSphere 5|
|Hypervisor Cost||$0||Platinum: $5000/server
|Management Cost||SMSD: $2,620/processor
SMSE: $1,569/4 VMs
|80VMs/6 CPUs/3 Servers
|80VMs/6 CPUs/3 Servers
Enterprise: $25,875 (an extra 3 licenses required)
So in essence what this means, is to be economical we start to scale out, which could increase our power costs, or we scale-up and upgrade to Enterprise+ licenses until we hit the same ceiling and then scale-out once more which will increase our power costs. This will have a decidedly negative impact on those who want or maintain a green datacenter or those datacenters that are already at their max power thresholds but instead of worrying about this now, we need to worry about this a year from now. In the meantime, many will start looking at alternatives.
Prior to the initial vSphere pricing announcement customers were generally quite happy with the VMware product, and while many felt that VMware was being aggressive on the pricing front, felt that they were getting acceptable value in return for the premium pricing for vSphere 4. This same customers had in general no motivation to look around, consider and evaluate competing offerings as they were content with the existing ROI picture for vSphere.
The initial vSphere 5 vRAM pricing shook the incredible good will and trust that was felt towards VMware. These changes reduce the initial sticker shock that many people experienced when they ran the vSphere 5 licensing numbers. In particular the provision that no single VM can get charged for more the 96GM of vRAM no matter how much it is actually using, will help assure people that virtualizing very large business critical applications will not become prohibitively expensive.
However the real issue for VMware and the customer base is whether this new pricing (even in its modified form) will create an incentive for customers to look at competing platforms and perhaps either choose something that is less expensive but “good enough”, or choose to be in a multi-vendor environment for virtualization platforms. Most enterprises purposely choose more than one vendor at each layer of the stack for infrastructure software so as to maximize negotiating leverage with the vendors involved. Finally, if this is how enterprise virtualization is going to go (towards more than one platform per customer) then this will have a dramatic impact upon the management and security strategies for all of the vendors in the ecosystem.
Share this Article:
Latest posts by Edward Haletky (see all)
- Finding your Sensitive Data to Protect - March 27, 2017
- Scale and Engineering - March 23, 2017
- SDS and Docker: The Beginnings of a Beautiful Friendship - March 21, 2017