Running with One Shoe
Cloud technology isn’t enough; we need to bring cloud economics as well to the enterprise

Last weekend was the first of three weekends on which an H1N1 flu shot day was held in Santa Clara, California. There was such an overwhelming demand for this shot that the net result was a massive shortage and long lines of residents eagerly awaiting shots for themselves and their children.  So here I am spending my entire Sunday standing in a line that was literally a half-mile long. I got in around 9:00 a.m. and exited around 4:00 p.m.  Some folks even arrived as early as 4:00 a.m. to grab a spot in line. It rivaled any Black Friday or Pearl Jam event.  

On the first weekend, the County ran out of vaccine shots pretty quickly and had to turn away folks who had stood in line for quite a long time. On the second weekend, evolution kicked in – they allocated more vaccines, ensured a count of folks during the day, and issued tickets to manage the count. Once the count reached 5000 (the max capacity that day), they turned people away ahead of them actually spending time in line. Although this approach overall was an improvement, it didn’t address the root cause of the problem – completely avoiding the six to seven hour wait is what we really needed.

This whole incident reminded me of how enterprise IT (and, in general, business) deals with capacity management, or the lack thereof. Inevitably, there are events, some predictable (the Christmas shopping season in online retail) and some unpredictable (catastrophic events being covered by media outlets), that cause spikes in demand. Traditionally, as we all know, enterprises have adopted the “give-room-to-grow” architecture, which I think could also easily have been called “let me pay for 70% of un-utilized resources.” Obviously, the recent economic indigestion (mildly speaking) has changed all that.

Enter various dynamic capacity management techniques including virtualization, on-demand architectures, etc.  All are driving towards improved utilization among other things. But the fact remains that one can never really predict demand. In fact, IT decision makers need to build in unpredictable demand into their business and technology model without the cost outlay.  

So which business model has truly internalized this?

The cloud computing business model for one. A key component of the cloud model that one needs to realize is that seldom does a cloud provider build in upfront CAPEX. It doesn’t matter if their technology can bend it like Beckham, the core underpinning of their balance sheet stems from the fact that when customers pay, they pay their vendors. When their customers scale, they grow and in turn the vendors grow. Period.

So why isn’t something like this adopted for the enterprise?
First, DNA
o       This is changing – no doubt the recent economic wakeup call has helped, but beyond that I believe CIOs, CFOs and their organizations are beginning to internalize that on-demand pricing is as critical as long-term support in the new decade.

Second, LEGACY
o       Again, with the evolution of virtualization, and technology morphing to support an on-demand model, the footprint in enterprise IT is slowly but surely getting there. As an example, the number of deployments of Citrix XenApp that have been virtualized has increased more in 2009 than any prior year. In addition, desktop virtualization is forecasted to be one of the most widely-adopted technologies of 2010.

o       This is the hard problem, but also the most important to solve. Essentially, the core economic issue is not just how the technology is consumed (e.g., usage-based metering, etc.), but is dependent upon how the technology is procured, paid for and scaled up to align with business growth.
 Getting the economics wrong, but the architecture right, is like running with one shoe – it’s the worst of the options.  So how does enterprise IT ensure it is wearing both shoes?

 –          First, internalize that enterprise datacenter capacity is now inherently as unpredictable as the online or cloud models.

-          Second, drive for pay-for-utilization or, as we call it, “Pay-as-You-Grow” in the core design, choice and licensing of technology. Ensure that IT decision makers keep this in mind from the get-go.

-          Finally, close the loop on iterating through the financial “grain” (i.e., whether you pay per month vs. per minute; whether you pay per GB vs. per IO throughput). The key is to retain the flexibility to change the scope of Pay-as-You-Grow pricing based on usage over the year.

So how does one manifest Pay-as-You-Grow in a product?

Take the example of networking. Traditionally, application delivery controllers were purchased using perpetual licensing based on throughput tiers and functional segmentation (e.g., buy an 8 Gbps appliance with caching and an application firewall for $85,000). What if your current scaling need is only 3 Gbps, but you believe that seasonal or unpredictable spikes can drive this to more than 5 Gbps? In the new world of cloud economics and Pay-as-You-Grow pricing, this translates to:
-          giving customers the choice of going with a 3 Gbps model today, and the option to scale up to 8 Gbps on demand, as needed
-          without forklift upgrades
-          without service interruptions
-          and, all of this is delivered with a simple license-key based activation procured via a self-service interface with back-end accounting that is automatically processed

That’s cloud economics.

That’s the other shoe.

Bottom line – based on conversations I have had with various cloud providers and enterprise IT leaders, I am a firm believer in the theory that technology delivered and consumed in the enterprise needs to natively support Pay-as-You-Grow pricing. This is the only way that enterprises will be able to align with the cloud model. It’s no longer just about the technology architecture.
Running is a whole lot more fun with two shoes.