Cloud computing is one of the most overhyped phenomena to have hit the IT industry in a long time. It is a business model that definitely has its advantages. The trouble is vendors of all sizes and stripes are so desperate for a piece of the cloud action, they are willing to blur distinctions and fudge definitions for their own ends.
Their headlong pursuit has saddled cloud computing with so many misconceptions that it is sometimes difficult for customers to make informed business choices. ZDNet UK has looked at the most common myths, and debunks five of them here.
Myth 1: Cloud equals SaaS, grid and utility computing
The term 'cloud computing' has been hijacked by anyone wanting to make a service sound hip and interesting. Jumping on the latest bandwagon is a favourite pastime in the technology industry, but in this case it is creating confusion among customers, who are unsure what they should be asking for or what they're likely to get for their money.
So to clarify: cloud computing is a form of outsourcing by which vendors supply computing services to lots of customers over the internet. These services can range from applications, such as customer relationship management, to infrastructure, such as storage and the provision of development platforms.
The services are provided by massively scalable datacentres running hundreds of thousands of CPUs as a single compute engine, using virtualisation technology. That approach means workloads are distributed across multiple machines — which can also be located in multiple datacentres — and capacity can be allocated or scaled back according to a customer's needs.
Moreover, because applications are multi-tenant in nature — multiple instances of the same package that can be executed on the same machine — system resources can be shared among a large pool of users, which reduces costs.
Software-as-a-service (SaaS) is one of cloud's most recognised manifestations, but strictly speaking it is a subset of the category because not all offerings are massively scalable. This means that, while some SaaS offerings are cloud, cloud is not SaaS.
Grid computing, on the other hand, comprises a networked cluster of heterogeneous, loosely coupled machines. These machines work together to solve a single, generally scientific or technical problem that is either CPU-intensive or requires access to large amounts of data.
The clusters may be located either within a single organisation or form part of a larger-scale public collaboration between many organisations, but each machine runs software that apportions pieces of a program to be processed. This means grid does not provide customers with an outsourced service but rather a way of clustering disparate machines to harness idle compute power.
Utility computing, meanwhile, refers to the packaging of computing resources as metered or subscription services in a similar fashion to those provided by water or gas companies. As a result, cloud can be purchased in an utility computing format, but the one is not predicated on the other.
Talkback
"It's stupidity. It's worse than stupidity: it's a marketing hype campaign. Somebody is saying this is inevitable--and whenever you hear somebody saying that, it's very likely to be a set of businesses campaigning to make it true."
Whether you like it or not, RMS has the uncanny knack of being right.
I would take issue with the suggestion that multi-tenant applications means "multiple instances of the same package that can be executed on the same machine"; that's virtualisation, and doesn't require any particular magic at all. However it is grossly inefficient from an operational viewpoint - it means if I have 1000 customers for my cloud-based Exchange service (e.g.) I have to separately manage 1000 instances of Exchange.
A true multi-tenant application is a *single* instance application, albeit horizontally scalable on a massive scale, which is aware that there are multiple distinct entities using it and which is able to create very robust chinese walls between them, as well as effectively manage their resource utilisation. From the customer's perspective it's the same as virtualisation, insofar as it appears to him that he has exclusive use of the application instance, but from the service provider's perspective it scales much much better. There's only a single instance to manage, and I don't have to worry about how to distribute resources amongst the various instances. When I add more capacity its available to be shared by all customers (subject to whatever QoS policy I may apply to individual customers). These aspects are critical for keeping the costs down.
I don't think I said that multi-tenant was virtualisation, but thanks for your extended explanation anyway.
For cloud computing to be a reality doesn't the service have to be application blind? If I were a company with either a massively peaky application - a disaster information website, serving Olympics facts and figures or organising annual pilgrimages for example - or use of it was growing/declining very fast at an unpredictable rate then the idea that I could purchase server capacity as a commodity would be very attractive.
If I had to make changes to my application in order to do so this attractiveness might disappear?
It seems to me that cloud computing might be a right choice if a) I was setting out to build a new application from the ground up and b) that application was relatively simple with a largely undifferentiated set of uses.
Anything complex with legacy stuff, web stuff and other technologies - database, .net and so on - all mixed in (pretty much standard these days) might not fit the cloud concept very well, if at all?
Since the Network Cloud Service providers had to build buildings that had high-quality power and were physically secure, it made sense for them to begin to offer data center space for these emerging consumer application cloud services. In the early days it made more sense for Google to use Savvis to provide data center space. Today, companies like Twitter, Facebook and OpenTable continue to rely on data center services provided by companies like Savvis, NTT, Terremark, and others. Of course, anyone who becomes a student of the cost of computing comes to realize that the cost of power is a big driver. As a result, anyone who needs space for 100,000+ computers (e.g., Google, Amazon, Microsoft) is building data centers located near low-cost and reliable power.