A better way to understand cloud computing

Earlier Wednesday, I wrote about the consensus on the need for a cloud taxonomy that was reached by the participants of the Cloud Interoperability meeting prior to Cloud Connect last week. But a couple of cloud ontologies have come to light that provide a great starting point for taxonomy discussions.

They are very similar, yet they differ in some noticeable ways. Nevertheless, both serve their purpose admirably and are required reading for those considering common understanding in the language of the cloud.

The first was brought to my attention by John Willis via Reuven Cohen. It comes to us from Lamia Youseff of the University of California at Santa Barbara, and Maria Butrico and Dilma Da Silva of the IBM T.J. Watson Research Center. They bring us a relatively simple five-layer grouping:

(Credit: Lamia Youseff, et.al.)

This ontology starts with firmware and hardware as its foundation, eventually delivering us to "cloud applications." Along the way, the requisite terms for software as a service (SaaS), platform as a service (PaaS), and information as a service (IaaS) are categorized, as well as the more rarely used data storage as a service (DaaS) and communications as a service (CaaS), representing storage and networking respectively.

As Reuven notes, it was driven by a need to define a common cloud classification model to allow academic research to advance.

The second ontology is the work of my good friend Chris Hoff on his Rational Survivability blog:

(Credit: Chris Hoff)

The powerful thing about Chris' ontology--a work in progress--is the completeness of the stack, from facilities to hardware to software infrastructure to applications and services. He then maps it to several example concepts in what amounts to a very early taxonomy that meets this ontology.

Yes, it is much more complicated than the UCSB-IBM contribution, but it lays out classifications that many of us find somewhat intuitive and certainly very close to complete.

That being said, note the overlap between the two. SaaS, PaaS, and IaaS form the core basis of the cloud-centric classifications. In fact, the most striking difference in that regard is that Youseff's team classifies DaaS and CaaS as separate from IaaS, while Chris defines IaaS as containing all three (compute, storage, and network). I would tend to side with Chris on that point.

Both projects would accept feedback, I'm sure. Chris Hoff is looking for commentary on his post. (I have some minor adjustments to suggest myself.) I am betting that Youseff's team will be reading the commentary to Reuven and John's posts. And I would love to get your comments here on the relative merits of the two approaches.

2008 will be remembered as the coming-out year for cloud-computing services. I'm beginning to think that 2009 will be best remembered for cloud-computing understanding. I encourage you to get involved.

[botchagalupe]

fyi,, It's not a coincidence that Chris Hoff's "Great" taxonomy happened to fall on the same day as my original post. Chris credits his original inspiration from my earlier post this morning. All and all this is a great thing that Chris, Reuven and you have done today. I am glad in some small part I was sort a connector for all of this.

Thanks
John
johnmwillis.com

[AppleSuxLeo]

The Palm Pre IS "cloud computing" It is the first smart phone that doesn`t require syncing with a computer.
A perfect example of why cloud storage is best is when I built a new system and did a clean install of the OS. When I DL`d Skype , voila ! All my contact info. was right there when I fired up Skype for the first time.
Then I thought , wow , the cloud is the way to go.

[knoxi1712]

I prefer the first model for simplicity. The second model, however, does have the notion of management and orchestration which is a key component of a cloud ontology. Although there aren't many solutions in the management and orchestration space yet, this category will grow quickly as cloud computing becomes more prevalent and companies look to manage open cloud deployments in multiple clouds. It's a broad space which will cover security, multi-cloud management, virtual machine orchestration, cloud marketplaces, etc. Existing vendors such as CohesiveFT, Elastra, Rightscale, Hyperic and Skytap currently fit in this layer. I proposed adding a vertical bar to cover management and orchestration to the side of the first model yesterday: http://cloudcastblog.com/2009/01/cloud-ontology/. With this modification, I think the first model is the simplest and most useful representation of the cloud stack for now.

-Ian

[Lyouseff]

Thank you James for sharing the UCSB/IBM model with everyone, and for the good points you make here. I have some points to add to the discussion, which I have sent through a post here: http://groups.google.com/group/cloudforum/msg/09012120e0115644?hl=en. However, I am including here a shorter version of the email.
Basically, I would argue that cloud-management applications -- those providing multi-cloud management, VM orchestration, fault-tolerance....etc, are just untraditional examples of cloud applications. In the end, they provide services to other cloud applications, and use services from other layers of the cloud like PaaS and IaaS.
The second point I want to highlight here is the importance of the simplicity of the model. I strongly believe that the simpler the model, the faster it will be adapted in both business and academic circles.
Finally, another important characteristic of the final ontology model would be to capture composability and interoperability between the different layers/components. We were careful in designing this diagram to show that, if two layers share a horizontal line, this depicts that the higher layer can be composed of the underlying layer in the stack. We explain that further in the paper. Further, we argue that if we can clearly depicts the relation between the layer (through composability), then we can realize interoperability opportunities between different offerings. More on that in my email response. I am not -in any way- saying that this ontology is complete, but i just wanted to highlight some of our thoughts when we designed it...in the hope they might be useful for the community.
Thanks again for a very fruitful discussion,
-- Lamia Youseff

[technogator]

I really appreciate the granularity that Chris provides, as I believe from both an academic perspective and to help illustrate the value of Cloud Computing. It's key to understand that the "Cloud" is dependent upon the Facilities and Connectivity layer in order to deliver value to the end customer. Service providers like Terremark and others are continuously communicating this as a key component to successful delivery of services within the "Cloud".

[glamourati]

It is a waste of time to try to lay out how it will all work, and that makes your article a very good one. This same attempt to "perfectify" it from the outset has always been there as new standards evolve. The reality is, nobody gets it right on the first try and users and the market will ultimately choose what they want. What usually happens is one of two things. Either somebody steps in and says "enough is enough, here it is" and it is adopted because even the inferior idea that is out there in the market is better than the best one that is still in the drawer - see the ODBC story. Or, more likely the big fishes force their partners to do it their way - see Walmart. The standards will be de facto in a cloud environment as most systems are on the Net. Also, what is more important than use cases, UML, whiteboards, etc. is what the end-user experience will be. It has to be simple, open, simple and open - did I mention simple and open? Making systems hard is easy, making systems easy is hard. To the same, the front-end of the cloud will be the most important part because that is the part the end users will interact with. Do you care what happens when you get money from your ATM? SaaS's PaaS's will be the front-end of the cloud, but SaaS and current PaaS's are closed so they will just be one phase in the long run. The next phase will be Open Platform as a Service (OPaaS). OPaaS companies like Modbox http://www.sullivansoftwaresystems.com/modbox will put the needed front-end on the cloud "water for the cloud" if you will.