Comments on: Definition of Cloud Computing, incorporating NIST and G-Cloud views

By: katecw

katecw — Wed, 23 Jun 2010 15:11:48 +0000

I would say that there are two utility platforms appearing; the Linux-Apache-MySQL-Php/Python/Perl (LAMP) stack for open source users and the Microsoft .NET-MSSQL stack. The hardware underneath has already commoditised and standardised in the form of x86 virtual machines. No one in their right mind is creating applications that need custom hardware these days, well apart from those poor few who succumb to vendor-sales-speak!

There are few applications which, with a little intelligence, cannot be designed and built to be highly scalable and operate on any old x86 tin (virtual or otherwise). As you might expect, my preference is towards open source as the platform, and that certainly seems to be the choice for new entrants providing Web based software-as-a-service (SaaS), based on our customers.

As for removing all technology decisions, if you want to do that go direct to a SaaS provider. My advice, though, is to separate your software supplier from your hosting provider, for improved security and resilience (see my post on supply chain disintegration). That also means you don’t get locked in to the vendor, and can use open source software. That still doesn’t meant you need much technology experience – just don’t go for solutions that need any special tech. If they do, they are probably a rip-off!

By: Scott Butler

Scott Butler — Thu, 17 Jun 2010 08:44:46 +0000

Interesting. For years we have been operating in technology silos, delivering and tailoring the infrastructure (hardware, OS & Middleware) to meet the clients needs. However, we are now developing our technologies into utilities but I am aware that the terms ‘utility’ and ‘cloud’ seem to be used in the same context, you address this here and this is a message that I will be reiterating over the next few months I am sure.
The biggest issue we are facing (I lead one of the Middleware teams focused on application hosting) is the proliferation of third party applications; I would be interested to know how we can have a true ‘utility’ model when the business applications vary greatly in their infrastructure requirements. A key point of utility, and ultimately cloud, computing is to removed the technology decisions from the client side but this would mean creating a solution that is all things to all people and I have reservations that this is achievable.
I would be interested in hearing your thoughts on this.

By: Cathie Gibbens

Cathie Gibbens — Tue, 09 Mar 2010 13:30:07 +0000

I like the equation you put forward and the diagram of NIST’s definition – makes it simple and clear, which is what cloud computing should be.
Suppliers are recognising the need for interoperability, but as you say, this is not as strong as it should be.
Looking forward with interest to playing spot the difference on the next post!

By: Kate Craig-Wood

Kate Craig-Wood — Thu, 25 Feb 2010 12:08:32 +0000

@SamJ Good point re. few actually needing elasticity. That is why we (Memset) have still not gotten around to implementing an API for our Miniserver VMs – there is little demand. Most people a) don’t want to re-write their apps to scale resources, b) don’t need a lot of scaling and c) view the hosting element as a very low cost compared to the service being delivered thus are happy to pay the premium to have overspecced resources there to catch load spikes. The views are changing, slowly, but I personally don’t think Amazon EC2 / instances-on-demand are the right way, which is why we are developing a Better Way to be more efficient about burst resource capacity – but its not ready yet so can’t talk about it.

By: Sam Johnston

Sam Johnston — Thu, 25 Feb 2010 11:57:26 +0000

Some interesting issues raised here, though like last year’s cloud definition season, 2010′s too will go largely ignored by the masses (who are happy to equate “cloud” to “out there on the Internet” – which is not the same as cloud == internet btw). That’s fine by me too – the promise of cloud is in letting a mutually trusted third party take care of business – a necessary part of indiscriminate multitasking incidentally.

As for the NIST definition, I think it has its fair share of problems but it’s better than most and it comes from a reliable (hopefully) independent source. One such problem is the “requirement” for “rapid elasticity” ala EC2 – if I have, have always had, and will always have 100 users for example, then I can certainly benefit from cloud computing by deploying SaaS products to my employees but I couldn’t care less about elasticity. Similarly, being self-service is nice, but if I had to fax in a form and wait a week to benefit from economies of scale enabled by multi-tenancy then so what? And then there’s “broad network access”, which is just a weak way of saying “Internet”. Either it’s ubiquitous or it’s not.

Thanks for taking the time to write this up and here’s hoping we can beat the NIST definition into shape over time (though my experience getting heard by them thus far has been fairly ordinary).