In my post about the Public Vs Private Cloud debate, I pointed out to how the promoters of “only public clouds” idea use the financial component as a requirement in the very definition of cloud computing. In their quest to drive home their point about the economics behind the public clouds, these pundits, either knowingly or unknowingly, promote a myth about private clouds. They argue that private cloud is not a cloud because it doesn’t take a multi-tenant approach and the infrastructure is built exclusively for a single company/enterprise.
More recently, I came across this argument by Christian Verstraete of HP on Christofer Hoff’s blog. He argues that
Private virtual environments are existing for quite a while. They work great and have good return. Whether they are called cloud or not is frankly irrelevant. The benefit of the “cloud”, in the public sense of the term, is that it can leverage the needs of many customers, hence has a potential of great elasticity, and its datacenters have the potential to run at higher efficiencies. “Private Clouds” have less of that potential as they limit the leverage to one enterprise. So, in my mind it makes sense to talk about virtualized data centers or next generation data centers, or highly leveraged data centers or something similar. Calling them cloud blurs the boundaries and confuses people.
(Bold highlight is done by me)
The argument that the private clouds serve exclusively as a single tenant is a fallacy. In this post I will show how it need not be the case.
My first argument against this myth has been already put forward by many pundits who favor the idea of private clouds. The old fashioned way of allocating resources in the enterprise IT is done by allocating fixed resources to each and every department without any intra-enterprise sharing. If any particular department needed more resources due to growth, new infrastructure was provisioned to take care of their needs even though many other departments in the same enterprise were under-utilizing their allocated resources. This lead to a waste of time and money and it was not the optimal way to run the IT.
With the advent of Cloud technologies and the eventual maturation of private clouds concept, these enterprises could easily use the automatic allocation of resources and the associated elasticity of the clouds. Now IT could easily reallocate resources based on the needs of different departments without procuring new infrastructure. The enterprise IT could emulate the advantages of multi-tenancy of the public clouds by using the cloud technologies. Even though the economics are not the same as the public clouds, there are significant cost advantages to this approach. Any attempt to tie the use of private cloud to single tenancy is very simplistic at best.
When I attended the Under The Radar Cloud event at Mountain View few months back, I had the opportunity to listen to a pitch by a company called Zimory. I was surprised by some of the things they mentioned in their presentation but I dismissed them immediately because I was part of the “only public clouds” gang then. Zimory talked about a concept which, in the hindsight, I thought was a brilliant idea. They offer something called a Zimory Public Cloud where they act as a broker between those who need computing resources and those who want to rent out their excess resources. This is simply brilliant. Enterprises who use cloud based technologies can rent out their unused resources through Zimory Public Cloud and companies (whether it is a startup or an enterprise) can use these resources to complement their existing resources during sudden spikes in their needs. There are some advantages in this cloud that will suit the enterprise customers better than the public clouds.
- Guaranteed SLAs on certified datacenters
- It is possible to choose the supplier(s) of the computing resources. This will let enterprises clearly know who their neighbors are in a multi-tenant cloud system. This is almost impossible to do in a public cloud environment
- The very fact that these are unused resources from another enterprise makes it more secure than the regular public clouds
These are game changing advantages from the point of view of enterprise customers. The very fact that the use of cloud based technology makes it easy for them to participate in a resource trading system, like Zimory, is a good enough reason for private clouds to be called clouds. The very fact that these private clouds can act as a multi-tenant system by participating in this kind of a resource trading system, clearly, debunks the myth that private clouds are tied to a single tenancy. Well, most of the enterprises use it as a single tenant because that is how they intended to use it. However, the same private clouds have the flexibility to be part of a multi-tenant ecosystem if they wish to share their unused internal resources.
As I told in my previous post, there is no need to fight over names and terminologies. As I told in my tweet yesterday, let us leave the exclusionary politics to those who need it. As technologists, who are supposed to be more enlightened, we should be spending out time worrying about how we can innovate on the technology front than fighting over whether something is a cloud or not.
Krishnan,
Very nice! I agree in general with the advice to just “do it” and not argue over terminology as well as debunking the myth of single-tenancy in the enterprise.
I think if you look closely at large organizations you’ll find multi-tenancy is just as easy a fit in an “enterprise” as it is externally. Resources are still shared inside the enterprise, they’re just distributed a bit differently. The lines of demarcation between “tenants” are projects, departments, and lines of business instead of organizations – but they still exist. There is sharing and while the financial benefits may not be as great as in an “external” cloud, they still exist and allow an enterprise to “do more with less”, which is really what we want in the first place.
Lori
There are cases where an IT infrastructure group with an internal/private cloud could allocate compute resources and their costs based on a economic charge back model for what is actually consumed by a department or division. This seems preferable to the typical head count allocation model commonly used and despised across enterprises.
In fact our company was working with a major investment bank (unfortunately never implemented as they are no longer the same entity they were…) who were looking to take the payment for computing resources one step further. They were investigating an approach to allocating a shared pool of physical host computers through the use of future and spot market semantics. This approach treated computation as a fungible resource whose ownership could be freely exchanged amongst market participants, i.e., the various, departments and divisions of the bank.
Lori and Route_b, Thanks for the comment.
Lori, that is my point too. I completely agree with it.
Route_b, thanks for sharing the info. Interesting!!