I followed a link to this "short-sighted and selfish" view from Lori @ f5's Approaching cloud standards with end-user focus only is full of fail rant and have to say that as an independent consultant representing the needs of large enterprise clients it's not surprising that I should agree with you (representing the needs of end users in general) rather than a vendor.In summary, cloud computing is finally an opportunity to shift the focus from the vendor to the user, where it arguably belongs. Vendors don't like this of course (and anything they say on the subject should be viewed accordingly) and are doing everything they can to stake a claim in what is something equivalent of a gold rush. Only this time (unlike the dotcom bust) it's real gold we're talking about (not fools' gold) and a large, sustainable (albeit heavily consolidated) industry of "computer power stations" and associated "megacomputer" supply chains will result.
Cloud computing is a paradigm shift (like mainframe to client server) and attempting to document it all in one rigid "ontology" is a futile exercise, as evidenced by the epic failure of attempts to do so thus far. A birds eye view of the landscape is possible, but only in the retrospective sense. One of the great things about cloud computing is that it is user-centric - for once the end-user has an opportunity to call the shots rather than being told what to do by vendors.
My various efforts (writing the Wikipedia article, setting up the Cloud Computing Community and more recently working on cloud standards starting with Platform as a Service) have all involved looking at what innovation is taking place in the industry and determining the consensus. Now is a very good time to do so as well because there are enough data points but no de facto proprietary standards (though the EC2 API is worryingly close to becoming one).
I tend to take advice from vendors on this topic with a grain of salt because most of their input tends to involve pulling the resulting "open standard" closer towards their particular offering - Enomaly's Unified Cloud Interface (UCI) for example not only focuses on VM provisioning but goes so far as to include them specifically alongside Amazon and Google.
The user doesn't [need to] care about this level of detail any more than they need to care about how a coal-fired power station works to turn on a light. The whole point of the cloud is that it conceals or "abstracts" details that ultimately become somebody else's problem. Using the power analogy again, our "interfaces" to the electricity grid are very well standardised (2-4 pins and a certain voltage cycling at a certain frequency) and "The Cloud" needs similar interfaces (for example for storing data and uploading and managing workloads).
Once we have that computing will be quickly commoditised, which is every users' best dream and vendors' worst nightmare (except for the few, like Amazon and Google, who still have a seat after the computer industry's next round of musical chairs).
13 March 2009
Approaching cloud standards with *vendor* focus only is full of fail
So I was taking stock of the cloud standards situation and found an insightful article (Cloudy clouds and standards) over at ComputerWorld via a colourful counterpoint over at f5 (Approaching cloud standards with end-user focus only is full of fail), hence the title. I made a comment which quickly turned into a blog post of its own (and was held for moderation anyway) so here goes:
Subscribe to:
Post Comments (Atom)
4 comments:
I want to express my wholehearted agreement with your observation that the CCIF and many others tend to be a bit too low-level and technical in their view of the problem domain at the expense of the business case for Cloud Portability and Interoperability.
I'd even posted this opinion to the group myself last month, saying in part:
"Cloud Portability, Interoperability and Compatibility are necessary to mitigate risks and minimize financial losses to organizations moving to and from the Cloud, and to help protect and insure the continuation of their business operations when (for any reason) a Cloud-hosted solution must be relocated."
And..
"I'll have to admit to everyone here that, while it is often used in CCIF conversations, I can't really imagine the word "hypervisor" ever being uttered by a group of software architects and developers discussing Cloud Portability in the context of how to avoid a total rewrite when moving an application between Coghead, Salesforce and Intuit."
IMHO, taking an exclusively VM, Hypervisor, XMPP etc. approach does nothing to solve the problem at the level of organizations trying to enter, leave, move around and survive in the Cloud.
Hi Sam,
First, if your response was held in moderation too long I apologize. I try to keep on top of feedback and get them approved as soon as possible.
The problem seems to be the definition of "user". The "user" of cloud computing is not John Q. Public but the developer and the organization. The ability to move data from cloud to cloud is not the same as moving applications from cloud to cloud. The "end-user" cares about data portability, the "user" cares about application portability. There's a huge difference in those two needs and the latter is primarily going to be served by some sort of standards such as those CCIF and others are working on.
If the organization cannot move from one cloud provider to another, completely, then standards have failed them. And focusing on the "end-user" will lead there because there's so much more to an application than just its data.
If it's selfish and short-sighted to be looking at the "bigger picture" and trying to include all the relevant components when it comes to interoperability and portability, then so be it.
Lori
@Bloggenstein: You're right, the focus on hypervisors and associated junk is more damaging than anything else, which is part of the reason why I chose to kick off the Cloud Standards Group at the other end of the stack (platforms)... this is where I see the real value is and a few Amazon-style providers and VMware-style offerings (ideally some open source ala Eucalyptus) is all we need in this area. For portability the DMTF have already given us OVF so we're pretty much good to go.
Is it any surprise though when the guy who runs the show (Reuven Cohen) also runs Enomaly, who have been struggling with getting out a virtual machine manager product for the last 5 years (I say struggling because what they have achieved thus far doesn't implement basic functionality, is woefully insecure and doesn't work anyway). Based on the feedback I've received I know I'm not the only one who has a problem with this massive conflict of interest, and with the appearance of Enomaly ECP in their "Unified Cloud Interface (UCI)" architecture - only I'm the only one in a position to talk about it.
I'm particularly pissed now because Reuven blatantly lied about the moderation to calm his critics but is still filtering posts relating to standards "not invented here". It is stunning what this guy is willing to put on the line to ensure that his voice is the loudest in the room - truly amazing... I've never seen anything quite like it before.
Anyway we've got runs on the board already and we're still waiting for something useful from the CCIF after 6 months... seems pretty clear where the results are going to be coming from.
Sam, I wouldn't paint cloud data models (which is just another word for ontology) with the UCI brush. As William says here, there's a scope & intent problem with the UCI charter as-is. I also have technical issues with the model they've come up with, but in fairness I haven't had time to help them on specifics.
A data model is a good way of driving clarity. It doesn't have to be "right", though it should be modular and extensible, since we know we won't get it right, or wont' cover everything.
Post a Comment