The Private Cloud seems to be gaining traction within the halls of Enterprise IT and was a hot topic at Interop NYC 2010. But do Private Clouds exist? What are their shortcomings, and what does this mean for the Public Cloud?

 

I live in the Public Cloud world and spend my days with Force.com/Salesforce, Amazon Web Services and Google. I tend to agree with the their viewpoint that private clouds are not really clouds at all. However I was struck with how popular the idea of private clouds were to those at the Interop NYC 2010 conference. This seems to validate the opinion that companies are asking IT “What is our cloud strategy?” and they are flocking to the private cloud as it seems like a safer and more importantly “comfortable” option for IT to pursue. This demand is also driven by vendors who see this as a good opportunity to sell more hardware and management software to IT.   This trend seems to ring true with the global SI’s as Dan Elron from Accenture stated that the private cloud makes up around 80% of Accenture’s cloud revenue and is “growing like gangbusters”.

 

We all know that IT budgets are tight and are asked to do more with less and generally have a large backlog (usually years) of projects that are waiting in the wings. I can’t imagine that many IT departments are going to have the budget or time to fully embrace a private cloud initiative. If only 20-30% of IT has embraced virtualization (which has been around for ~10 years), how will they be able to turn that into a private cloud? I’m sure a few are diving in head first, but many more are probably trying private clouds in small ways or adding a level of self service over any virtualization they may have.

 

I heard a statistic during one session where the speaker shared that without virtualization you achieve about 10% utilization on your hardware. If you add virtualization it goes up to around 30-35%, and if you add a private cloud on top of that it goes to ~38% utilization. Why so low? The real issue behind these numbers is that almost all companies have some degree of spiky demand. Either there is more demand during working hours to handle the 9-5 end users, or more demand overnight for batch processing. Either way, virtualized, private cloud or bare metal, the hardware needs to be sized for that demand plus a certain amount of overage. That leaves troughs. Times where that capacity is not used or needed no matter how virtualized that hardware is configured.

 

The challenge with virtualization and private cloud isn’t just the technical aspects, it really centers around process and change management. Many companies use slow hardware provisioning and IT backlog as a mechanism to kill projects or ideas that should never see the light of day. Most IT departments are built around these delays and moving server provisioning to minutes from months can have unintended consequences. Sys Admin’s suddenly are trying to manage 5 to 10x the number of virtual servers if developers have easy access to provisioning their own servers. I’ve seen companies put processes and artificial delays in place to handle these types of issues and it lessens the advantages that can be achieved.

 

I’ve lived in the public cloud world for the last six years and certainly understand the benefits and shortcomings of the various leaders in this space. However I recently had an opportunity to speak on a panel at a Datacenter Dynamics conference around the topic of “The future of the data center” to provide a “cloud viewpoint”. I’ve been away from that world for awhile and I was struck how hard it really is to create and run a data center. You need to be an architectural expert (where to put cooling, raised floor, pillars, generators) and electrical engineer (how to power it, add redundancy, cooling, cabling) and a system administrator (server setup, configuration, management, backup). While my fellow panelists debated whether or not the mechanicals (fans, cooling, etc) needed to be on a UPS or not before temps hit critical mass during a power failure, I just thought about how much I love not having to worry about any of that with the cloud. At the end of the day, why do you want to have your own data center and worry about all of those types of details?

 

While talking to some of the Interop panelists who spoke during the private cloud summit  another trend emerged. It turns our that several of them don’t actually have a private cloud, but “they are working on it”. No one was able to articulate what it really takes to build one beyond using some of the newer offerings from VMWare or other vendors plus a lot of custom development. There is no clear answer today and it is made worse by the “Cloudwashing” going on by all vendors who are quick to state the cloud benefits of their particular solution.

 

So what does all this mean for the future of the private cloud? My prediction is that it’s future is bright for the short term. It “feels safe” for IT, is being pushed by vendors such as VMWare and IBM and is validated by global SI’s who see it as a great vehicle for long term consulting contracts. Longer term I think everyone will realize that it has limited benefits and but it will be a stepping stone toward the public cloud. As Allistair Croll put it during his closing remarks, the future is the Hybrid Cloud.