Monthly Archive

I gave about a 40 minute presentation on cloud computing at the Minneapolis VMWare Users Group meeting yesterday.  I was pleased at the number of participants in attendance.  My goal was to deliver a cloud computing presentation that was compelling, if not a bit controversial. I felt like I was going to meet the emperor to tell him he has no clothes.  Most emperors wouldn’t react kindly to this sort of news from their loyal subjects. I wasn’t sure if the participants would embrace some of my ideas or chase me out of the room.

I started the presentation with a basic overview of how I define cloud computing.  And I pointed out that most every vendor defines cloud computing differently — and usually in a way that specifically benefits the vendor.  I apologized in advance for picking on Vmware because I really like the company and their technologies.  I’ve used Vmware for years and I’ve gone through the VCP training. Plus, they throw fun parties at their Vmworld events.

I then started to talk about some of the myths and misconceptions surrounding cloud computing. It was at this point that I admonished my marketing people to cover their ears.  Because we’re as guilty as anyone else when it comes to hyping cloud computing. Sometimes we’re forced to call solutions “cloud computing” because that’s what other solution providers are calling them.  Solution uses a couple Vmware ESX hosts? Oh, that’s a private cloud. Computer is hosted in an Internet datacenter? Computer + cloud = cloud computing.  I digress.

The first myth I talked about was that cloud computing does not require virtualization. I saw some furrowed brows in the room.  My marketing people were looking at me like “have you lost your mind?”.  You don’t stand in front of  Vmware diehards and say that cloud computing <> virtualization.  But it’s the truth.  Cloud computing isn’t a technology as much as it is a business model.  It’s a business model that leverages outsourcing, computing resources, and business processes.  There’s no rule that says cloud computing requires virtualization.  Cloud infrastructure is abstracted from the perspective of the end user.  Virtualization technology is a useful tool to enable that abstraction but not a necessity.

I’m sure there are plenty of SaaS companies that don’t use virtualization or limited virtualization in their infrastructure. I know because some of them live in my data centers.  I can think of at least one hosting provider that has an IaaS cloud solution that doesn’t use virtualization. Virtualization offers a way to slice and dice computing resources into smaller units to increase resource efficiency.  But it isn’t a requirement for a cloud computing offering.

Next myth I dispelled was that cloud computing doesn’t require HA – meaning, it doesn’t require high availability at the infrastructure layer.  Vmware and other commercial hypervisor vendors really focus on capabilities like HA and vmotion at the infrastructure layer.  These technologies are designed to make applications more resilient.  But the early cloud platforms never incorporated these features and thus developers had to rely on other techniques to build reliability and scale.  Cloud scale applications scale horizontally rather than vertically so infrastructure availability isn’t that big of an issue.  Developers can leverage load balancing in the cloud to scale web applications.  They can take advantage of db sharding, NoSQL, and Hadoop to perform resilient data processing on the back end.  In the cloud it doesn’t matter if individual pieces of infrastructure fail because high availability is accomplished at the software layer.

I argued that Vmware was not the leading hypervisor technology in the cloud today. Now I admit that’s entirely debatable and depends on what you consider to be cloud computing.   I clearly don’t think that every box with a copy of Vmware ESX on it is part of a cloud computing platform. I’m also looking at this more from the perspective of IaaS versus SaaS. If you look at the IaaS public cloud landscape (which is clearly the most visible) the Vmware hypervisor is getting its butt kicked by Xen.  Amazon owns about 50% of the public cloud marketshare according to industry analysts like Tier1Research.  Amazon uses Xen.  And then you have a number of smaller cloud players like Rackspace and others (including our own ReliaCloud).  Most of the smaller cloud players use Xen as well. In the public cloud Xen wins in a rout.

Well what about the private cloud?  I pointed out that the existence of a private cloud behind the walls of the enterprise is up for debate.  The notion of an on-premise private cloud sort of contradicts the whole cloud computing-as-a-business-model concept.  One could say this is like going to Home Depot, buying a generator, and calling yourself a power company.  At some abstract philosophical level you might be right.  But that doesn’t mean your neighbors will knock on your door to buy power from you.  I believe in private cloud as a single-tenant hosted deployment model.  I think the idea of an on-premise private cloud is promoted by vendors that need to sell you more stuff. Of course that’s my perception and I’m as biased as any other cloud vendor.

I received polite applause from the audience as I wrapped things up.  I really couldn’t gauge their immediate reaction.  Over the next day or so I received a number of messages from participants thanking me for the presentation.  A few people told me that my presentation really caused them to think differently about their in-house Vmware deployments and the way they viewed cloud computing. I really appreciated the feedback from everyone I talked to!