Pricing Elastic Application Delivery

One of the claimed benefits of cloud computing is that you can scale up or scale down applications based on demand. Autoscaling is supposed to be automated and on demand, which is great in theory, and can even be done in practice, but automatic scaling ignores a related but important issue: licensing. Citrix announced their Burst-Packs for Netscalar at their user conference. It's an interesting concept, but it's inelastic.

The Netscalar Burst-Packs are 90 day licenses that allow Netscalar customers to take existing Netscalar hardware and virtual appliances up to full capacity for a 90 day period. The license is priced at 35 percent of the difference between the currently licensed capacity and the burst capacity. Citrix selected 90 days based on customer demand. A representative said "We chose 90-day packs based on a lot of customer input. Internet and cloud customers loved the idea for use in seasonal and promotional kinds of spikes, most of which have a window beyond 30-60 days. For enterprise customers, the main use case is instances when traffic is unpredictable. We considered having multiple lengths, but everyone we talked to thought 90 days was about right, we opted to keep it simple."

Simple is good, and let's face it, this whole cloud thing with elastic pricing is pretty new to IT. Fixed pricing is at least predictable both for the customer and for the vendor. However, I don't think that fixed pricing like Citrix is doing with Burst Packs are going to last. At least, I hope not.

A 90 day block of time is good when you have a predictable event that you know will last for a while, like the holiday shopping season which runs from November through January. But in Citrix's own words, the enterprise may have more unpredictable demands that run less than 90 days. For example, universities have known periods of extreme activity on their registration systems during the start of the fall semester and the weeks while class registration is open. These are times when an application delivery controller can do wonders to cost effectively allow a university to scale upward as needed. At other times, demand low very low. A 90 day window is too long.

I think Citrix's Burst-Packs are a first step towards elastic pricing for network services. It's a different revenue model for vendors. A university that needs to burst three times a year should just pay for the oversubscribed license and leave it at that. However, that is wasteful and doesn't meet their demand curve. I imagine other vendors might start offering demand base pricing in more granular units (I would), and that becomes part of the selling point to potential customers. "We can do all this app delivery stuff for your and the cost will be low and incremental." It's a win-win.