Why all private clouds being promoted these days are just Infrastructure-as-a-Service (IaaS) clouds?
Why we have several vendors that offer hardware virtualization platforms for on-premises IaaS clouds and nobody that offers something similar for private PaaS and SaaS clouds?
On average, the commonly shared long-term vision for cloud computing is about a world where off-premises facilities host 100% of business applications and where nobody, except cloud providers, has to care about the underlying complexity of operating systems, physical servers, storage arrays, networking, security, and everything else comes to mind when we thing about our existing data centers.
Actualizing such vision requires addressing a number of technical issues (related to security and compliance, orchestration, billing, etc.) and changing the mindset of individuals inside corporations (the concepts of “physical ownership” and “trust what you see” are two of the strongest in human mind).
Because of these challenges the on-demand world above is still, say, ten years away and the industry is approaching it with little steps.
The first one definitively is IaaS private clouds, but looking forward this is just a transitional phase, pretty much like hardware virtualization is a transitional phase to application virtualization.
Before moving all their business outside the company’s walls, customers will likely first embrace IaaS private clouds, then PaaS and SaaS private clouds and ultimately SaaS public clouds.
So why vendors like Google, Zoho, Salesforce or even Facebook aren’t interested in delivering an on-premises version of Google and Zoho Office, Salesforce CRM, or “Corporate Facebook” today?
Why Microsoft just plans to sell subscriptions for a hosted version of Office 2010 rather than offering the web version to the companies that want to host it internally?
Someone may argue that this is not the business model of these companies, and in some cases that’s an entirely valid objection, but is it possible that none of their competitors wants to bet on the idea of private SaaS?
A possible explanation for the lack of PaaS/SaaS private cloud offerings is that the existing products rely on ad-hoc infrastructures that are fairly expensive, complex to replicate and definitively not market-ready.
The underlying grids that power them may be made of custom-designed physical servers plus recompiled Linux kernels plus proprietary agents (like Google does), or built by nesting together multiple virtualization technologies, glued together by daily updated fabric controllers that should be considered more a work-in-progress rather than a production-ready piece of software.
This is where VMware may want to be.
If the vendor can provide an out-of-the-box infrastructure for IaaS/PaaS/SaaS cloud computing, SaaS vendors adopting it can lower the cost of designing their data centers, and most of all they can offer on-premises versions of their products more easily.
Such out-of-the-box infrastructure should have abstraction layers that can be turned on and off on demand to address the needs of different customers: cloud-aware hardware virtualization (vSphere) for those that just want IaaS clouds, cloud-aware application frameworks (Spring) on top of it for those that want PaaS clouds, and white-labeled, cloud-aware web applications (Zimbra, Redis, and others to come) on top of it, for those that want to offer SaaS clouds.
If Facebook internally moves on such VMware infrastructure, and, say, a Fortune 500 moves on the same infrastructure, how much easier is to sell and deploy “Corporate Facebook”?
And if the customer doesn’t have the infrastructure yet, how convenient is for both VMware and Facebook to sell a bundle?