VMware recently announced a change to its software licensing model that caps the number of cores supported in the CPU, effective April 2, 2020. Why did VMware do this? What is the real impact to IT? And what does this mean for the VMware ecosystem and competitive landscape? The next few paragraphs will try to tackle these questions.
First – the details
Nearly all VMware software is licensed on a “per CPU” basis (Enterprise PKS and NSX Datacenter are the most prominent products licensed per core). Though the company flirted with the idea of “consumption based” licensing some years back, it quickly reversed course due to market push back.
Fast forward to a market where core-based licensing is normal, if not the norm, and the company looks to be moving to more of a “hybrid” model. In this new license model, VMware will still license per CPU, however, the number of cores supported per CPU will be capped at 32. Anything over 32 cores will require a second CPU license. If a second CPU is used, a separate per CPU license is required, with the same conditions (over 32 cores requiring a second license).
The above example from VMware’s announcement shows a simplified licensing chart. But what if the server on which VMware software is deployed has two 48-core CPUs? This would still require four CPU licenses as each license is tied to the physical socket.
Second – what does this mean for IT organizations? And what should IT organizations do?
Frankly, the vast majority of IT organizations will not be impacted by this licensing change today. The average server in the datacenter ships with less than 20 cores per CPU. However, as IT organizations look to deploy servers with more densely populated CPUs, such as AMD’s 2nd Generation EPYC Processor (which can scale to 64 cores per socket), there will surely be re-evaluation.
A suggestion for IT organizations that have begun to deploy EPYC in the datacenter: don’t panic and don’t immediately abandon this strategy. Rather, revisit the reasons for choosing to deploy these servers in the first place. Does this licensing model render those reasons inconsequential? Is there a deployment model that allows an organization to stay this strategic course and minimize the cost impact? Given the relative low cost of hardware (compared to the software solution stack), the answer to this latter question should be “yes.”
Another thing to consider is that VMware is giving customers until April 30 to purchase licenses for >32 core CPUs, which will be grandfathered and therefore not subject to the new license model.
Third – Why is VMware doing this?
First and foremost, dismiss the conspiracy that Intel is somehow behind this VMware decision (the company will be impacted by this licensing soon enough). There’s no “deep state” in the technology world that is trying to prevent AMD from finding success with EPYC. Rather, this is VMware responding to what it believes is a potential loss of licensing revenues due to the density of compute platforms.
As stated earlier in this article, if core-based licensing is not the norm, it is certainly normal. Both in terms of software licensing, software subscriptions and cloud services. Microsoft moved to core-based licensing years ago for Windows Server, SQL Server and other products. Likewise, database giant Oracle has been core-based for many years. Why wouldn’t the company that has upwards of 75% virtualization market share adopt such a licensing model? In the era of subscription-based software and this “as-a-Service” market, this signals VMware’s move to better compete.
Fourth – How does this impact AMD and Intel?
It is easy to look at AMD’s 64-core EPYC processor and leap to the conclusion that this licensing shift will majorly impact AMD as it ramps in the datacenter. However, the reality is the vast majority of EPYC-based servers are being sold with far less than 64 cores (or even 48 cores). These extremely high core counts tend to be deployed in support of high-performance computing workloads and applications with similar performance requirements. As a result, the real-world impact of this licensing model change should not be significantly impactful to AMD’s server business.
Likewise, the impact to Intel is insignificant. Even as it moves to a new CPU generation with greater core density, the mainstream virtualization parts will continue to be in the sub 32-core range.
Lastly – what does this mean for the future?
The impact of “the cloud” on IT is significant, making “as-a-Service” the norm in consuming hardware, software and solutions. While this licensing shift from VMware is not so drastic, it certainly enables customers to begin the transition of consuming its products more like these consumption models that are becoming so prevalent in IT.
I suspect that over time, VMware will continue to evolve its licensing model to more closely align to market dynamics and demands. In the end, this will ultimately benefit customers that find “as-a-Service” to be the “easy” button for providing IT services to the organization.