IT Management and Cloud Blog

« | Home | »

Does a Cloud Have to be a Grid?

By John | May 16, 2008

Honestly I am not sure.  Are Mosso and IBM’s Blue Cloud grids? My guess is no.  Therefore are they both still clouds.  I would say yes to the former and no the latter.

I have tried to avoid Dell’s Blog about cloud computing because I liken it to the “Fox in the Hen House.”   I am pretty sure they are last in line of people who are rooting for “Clouds” to succeed.  However Cris Sears another Atlanta “Cloud”enthusiast pointed me to the Dell blog link yesterday.  In the post the author makes some bold statements about what is a cloud definition.  He claims that it first must be a grid and then he goes on to say that if there is manual intervention it is not a cloud.  I pointed out in the comment section of his post that statement takes Amazon’s EC2 out of the “Cloud” running.

I would love to hear others opinions on the aforementioned question – Does a cloud have to be a grid?

Topics: amazon, aws, cloud computing, dell, ec2, ibm blue cloud, mosso | 4 Comments »

4 Responses to “Does a Cloud Have to be a Grid?”

  1. Wayne Pauley Says:
    May 19th, 2008 at 1:23 pm

    John,

    I’m not sure – and here is why.

    On the one hand I used to think of the Cloud as the “offer” and the Grid as the Technology. Lately however in my research (e.g. IEEE and ACM) I’m seeing ambient clouds, p2p clouds, etc. which are not the traditional Grid at all.

    If we add Utility Computing to the mix – we have yet another name for this. NIST and other organizations have talked about Cyberinfrastructure for a while – which I would equate to the Cloud – Internet based service offerings (could be anything from Library of Congress online to Salesforce.com to a realtime weather data collection system). Grid is a specific (or several specific) architectures for performing work – which may require a lot of end-user setup and/or may be purpose built. Utility is the service offering – all packaged up (send it work – out comes result). This is where I am now but I’m open to a much clearer and better set of definitions.

    One document that I recently came across is a Taxonomy for Grids (Venugopal, S., Buyya, R., & Ramamohanarao, K. (2006). A taxonomy of Data Grids for distributed data sharing, management, and processing. ACM Comput. Surv. 38(1), 3.)

    Venugopal, Buyya, and Ramamohanarao describe four models of “data Grids” – Monadic (centralized), Hierarchical, Federation, and Hybrid. They suggest that hybrid is the most likely in commercial production. They also do a nice job of describing the data transport and replica taxonomies inside the Grid which is useful when considering BC/DR, data locations, security models, etc.

    Their research is fairly recent – but the world of Cloud/Compute has progressed dramatically since then. Much of the basis of their work is tied to scientific/scholarly content. I think the upside is that it provides a fairly normalized model for newer instantiations and a great baseline for discussion and terminology standards (as opposed to marketecture).

    Wayne

    Wayne

  2. Roland Judas Says:
    May 22nd, 2008 at 6:16 am

    Hi John,
    I took the ball and wrote some lines in my private blog about your question. Just to sum it up I define grids as beeing more like an architectural pattern, while clouds are a modern computing paradigm.
    Because I prefer simple things I would love to come back with a new formula saying
    ‘Clouds = Grids + Internet’
    but I’m convinced that this is not _that_ easy.
    Even the Mosso.com guys seem to have a very individual definition of ‘clouds’: See http://www.mosso.com/vs_compute_clouds.jsp, where they compare their ‘cloud service’ against other ‘cloud services’ and other ‘typical hostet servers’. Am I the only one who gets confused?

    What interests me in your post, is your quotation of Jimmy Pike (Dell) saying ‘if manually intervention is needed, it’s not a cloud’. If this was true, we (my employer and me working on real IT automation) would be very happy. Everyone going for clouds would need our technologie (see our blog @ http://www.hcboos.net).
    Regards
    Roland Judas

  3. John Says:
    May 22nd, 2008 at 8:07 am

    Good stuff guys.

    Roland… no doubt Jimmy Pike was wrong and there will be great opportunities for cloud automation. I will check out your blog.

    Thanks

  4. Michael Sheehan Says:
    May 22nd, 2008 at 4:07 pm

    John,

    That is a great question. When we started developing GoGrid (http://www.gogrid.com) a few years ago, the “clouds” were high in the sky and barely seen, meaning, people were abstractly talking about it but it didn’t have the buzz like it does now. Grids, however, were “hotter” and thus, we named our product GoGrid. But it wasn’t just because of that. GoGrid is built on a grid of computers, with Xen virtualization, and nodes running the different instances of GoGrid. I have gotten into several discussions about whether it is a “grid solution” or not. The debate continues, however when you look at it more closely, we offer “control in the clouds” so to speak.So, essentially marrying a variety of things, cloud computing (we are heavily compared against EC2 in many ways: on-demand and scalability) with grid computing (more “under the hood” stuff than truly visible to the end user) plus utility computing (with metered pricing and virtualization).

    But to answer your question, I don’t think that a cloud has to be a grid. It is with GoGrid however. And I agree with your comment that the author did contradict himself. You DO have to be able to manage or control your instances in the cloud. There are some services that can be self-managed, but completely unmanaged? I don’t think so.

    Thanks, as always, for your continued great insight into this industry,
    Michael

Comments