VMware signaled a profound change in pricing as part of its announcement of vSphere 4.1. This move dovetails with Cloud service provider trends where VM-based pricing can be easily passed on to customers. However it sends a wakeup call to those customers facing VM sprawl with limited transparent chargeback capability - i.e. most Wikibon users.
Over the long term, this move will encourage better management practices related to VM provisioning and better resource utilization by providing incentives to create more balanced data centers. However in the near-term, users face hidden costs from VMware's new pricing methodology which warrants immediate attention to avoid surprises and CFO backlash. We strongly recommend users understand the new pricing model and implement at least a "showback" model if not a full chargeback system so that the business, not IT, pays for the consumption or possible over-consumption of virtual resources.
VMware made the following moves:
- VMware decreased the pricing for entry-level kits, making VMware more competitive with Microsoft Hyper-V and helping it to gain greater traction with the small business marketplace.
- VMware is moving from processor-based pricing for all VMware management products with the exception of vSphere. This will allow VMware to increase revenues for products that are clear market leaders.
vSphere 4.1 has a significant amount of additional function, and there is every reason to implement this release in a timely fashion. However users need to understand that the pricing changes mandate a change in the way that VM is managed. The current approach of many Wikibon users is to place as many VMs as possible on the latest hardware. This practice should be moderated in our opinion and incentives put in place to bring VM sprawl under control in the near term.
Lower Prices for entry level kits
VMware announced three kits designed for the small business market. These bring down the entry cost to about 50% of the previous level, and of course helps compete against Microsoft’s Hyper-V product. These kits are:-
- Hypervisor – Previously called “Free ESXi”, which consists of Hypervisor, single server ESXi, vSphere Client and Converter
- Price - $0
- Essentials – ESXi for 3 hosts/6 processors, 4-way vSMP, VC agent, VCB/ vStorage APIs, Update Manage and Thin Provisioning:
- $495 for six processors, $83/processor.
- Essentials Plus – All the elements of Essentials plus High Availability, Data Recovery and vMotion:
- Price - $3,495 for six processors, $583/processor.
Figure 1 shows the price per processors for the various vSphere kits and editions
Move to VM-based Pricing
VMware introduced a new method of charging for all VMware products except for vSphere. The reasons VMware gives for this strategy are that the Cloud (Private, Public or Hybrid) makes applications independent of the hardware on which they are running and renders the physical location of the gear irrelevant. It follows that VMware software licensing needs to map to an element that the IT organization can control, the virtual machine itself. In our view, the real reason for the change is VMware wants to maximize its revenue and limit its downside risks. Specifically, modern servers are packing so much power into each processor and the number of VMs that can be run is increasing so fast that the revenue trajectory with the current model is being threatened. The implication of this analysis is that VM-based pricing is almost certain to extend to all VMware for-pay software components in the future.
Overall, the move is sensible from both VMware's and the user perspective. This will force customers to tailor VMs to the exact business requirement and enable customers to set up an internal VM service that is comparable to external VMs provided by Cloud service providers. It also gives VMware an opportunity to quietly increase revenue from the management products, which are much stronger than those from Microsoft’s Hyper-V ecosystem. Customers that do not respond intelligently to the pricing changes will be writing unnecessarily large checks to VMware.
The VMware management products that will be placed under this scheme are:
- Site Recovery Manager,
- vCenter Configuration Manager,
- vCenter Application Discovery.
VM licenses will be billed on a Managed VM basis. The number of Managed VMs is calculated on an 12-month rolling average. This looks initially like a good deal – you don’t pay for additional licenses until a year later. But it also means that you could be paying for the sins of last year for many years to come if VM sprawl is not managed correctly.
Impact of Not Managing VM Sprawl
Figure 2 below looks at a three-year scenario where the VMs are allowed to rise from 100 to 200 in the first year, and then rise out of control to 400 in the second year. In year three they are brought under control, reducing the population to 300. This is shown in the blue line in Figure 1.
The red line in Figue 2 shows the Cumulative charged VMs, which are based on the maximum of a 12-month rolling average. This shows that even after the decision to bring the VM sprawl under control, there is still a commitment for additional VMware licenses.
Figure 3 shows the VM counts at the beginning and end of each year. Even though there was a decrease in managed VMs from 400 to 300 during the year, an additional 63 licenses had to be purchased because of the commitments made in the previous financial year.
Eliminating VM Sprawl
VMs have been very easy to provision, and the time for the business to provision has dropped from days or weeks in a physical environment to hours or even minutes in a virtual environment. The ease of creating VMs has led to VM sprawl.
The new VMware pricing model means that VMs must be managed far more closely. In particular the expected costs for both the initial project and the costs of full deployment should be included in the financial case for the VM. This will add to provisioning time; it is important for IT that the business own the financial approval process. This can be done by providing a self-service portal for the business in the same way that VM service providers do, except that the credit card is replace by charge codes and approval numbers. Approximately 15% of Wikibon users report that they have fully transparent chargebacks in place that could handle such a methodology. The risk to IT organizations without good chargebacks is that the business will continue to demand VMs, and IT will end up footing the bill.
In addition, the funding of VM software function needs to be managed closely. At least a "show-back" mechanism needs to be in place to let the businesses know what has been approved, and ideally a charge-back mechanism should be implemented similar to that implemented by VM service providers. Equally there needs to be strong controls to ensure that VMs that are no longer needed are deleted.
These changes will enable executive management to understand the efficiency of internal IT by directly comparing these costs with best-of-breed service providers. If the internal IT cannot be as efficient and there is no business security or legal factors that impel running on premise, outsourcing becomes an easy business decision.
Action Item: A CIO imperative for 2010 is to implement controls on VM creation, the software function that is provided for each VM and VM deletion. CIOs should be looking to implement self-service and charge-back mechanisms similar to those in place from VM service providers. If internal VMs cannot compete with those from service providers, they should be outsourced.
Footnotes: VMware Chargeback Software