SDN Licensing: Please Keep It Simple

As SDN products come to market, networking vendors are looking at new ways of charging customers. The talk is about consumption models, tiered licensing and variable costing. I'm asking vendors to use common sense to design pricing plans that are easy to comprehend and consume.

Vendors should recognize that early deployments of SDN will largely be performed in test beds, proof-of-concept labs, or skunkworks projects with little or no funding. Vendors would be smart to offer zero cost for pilots and early adopters to help spur these kinds of deployments.

Testing lets engineering teams fulfill business requirements to determine business fit, competency and service impact. In sales speak, this handles the "technical objection" phase in the sales cycle.

As a network architect, I regularly attend meetings where a project is reviewed, costed and analyzed in less than an hour. Each representative at the table gives a number, and the group decides whether to proceed to the next phase of analysis. In less formal settings, I discuss projects with colleagues who invariably ask for ballpark pricing of the product.

Five years ago, I could reliably predict networking costs using some rough calculations. Today, vendors have moved to layered licensing schemes where features are extracted into complex matrices that have accelerated price increases for non-obvious features. Today, I can't estimate pricing at any level of confidence. I must initiate a work package to determine pricing, research product licensing, and engage resellers and vendors.

[We don't know how much VMware's NSX costs, but we do know how the company plans to charge for it. Find out in "VMware Mum On NSX Pricing, Touts Licensing Plan."]

The outcome? Architecture work costs more, delivers less and ultimately leads to a devaluation of the design review process. Architecture work becomes a pointless exchange of sales engagement, quotation and research into mostly stupid licensing schemes that have no benefit to the customer.

Marginal and mid-sized projects wither and die for lack of simplicity to execute. Project initiation and execution is extended. The whole cycle is negative reinforcement of change in product adoption.

For SDN to succeed, vendors must develop licensing that can be explained in less than 10 minutes. For example, consider the following scenario for hardware and software:

200 servers needs 5 leaf switches at $7,500 per unit

The advanced switch feature license is 25% markup, so approximately $9,000 per unit

Each 10GbE port needs an SFP (the server team can buy the cable) at $500 for a total of $100,000.

Leaf switches are $145,000

5 leaf switches need 3 spines at $12,500 each, for $37,500

Total Hardware Spend = $182,500

200 servers need virtual networking licenses at $500 per unit

100 of those servers need advanced virtual networking features at 50% markup

100 servers at $500 each plus 100 servers at $750 each

Total Software Spend = $125,000

With this approach you can rapidly determine whether a project is viable. No time wasted by the reseller in pointless quotations. No time wasted with the vendor arguing over complex feature matrices and license requirements.

OpEx discussions are also a big deal in design meetings. For some companies, this is the only time that cost of ownership is considered, but it can be a deal breaker as a very early stage in the discussion. The support cost should be mental arithmetic based on the purchase price.

Don't Be Like Microsoft

Open source technology has created certain expectations in IT that are opposed to standard business practices. Anyone can download and install an open-source application or platform and evaluate it.

There is a resource cost associated with the testing and evaluation, but this cost applies equally to commercial and open source products. Indeed, a commercial product costs much more to evaluate because it requires engagement with a reseller or vendor staff who will want to evaluate the opportunity and "add value." Conducting an initial inquiry for a commercial product will take 10 to 20 hours minimum, and those hours have to be billed/charged/costed to the business.

Networking vendors must be careful to avoid the beast that Microsoft has created with its licensing scheme, which is now so complex the company has a large team of "licensing experts" and an entire certification program to support it. This approach provides no value to the vendor, reseller or customer, and Microsoft's reputation suffers because of its bloated bureaucratic approach.

Keep the software and hardware licensing schemes simple. You will encourage rapid adoption and enable the business. Don't make it complicated.