|
Here's my definition, for what its worth. Cloud Computing is the separation of the computing resources from the client by the Internet.
This implies that, just like in a real corporate server environment, all the primary computing services are located in a server bunker, properly resourced and protected. It also means that, just like in a real corporate environment, users connect to these services by way of a data service. Where it differs however, is that in Cloud Computing, the data service is a free and open Internet whereas for the corporate, it is generally some point-to-point dedicated channel from the user to the server (note that I'm choosing to not distinguish between corporate in-house and co-lo data centres – really it doesn't much matter).
So, what does Cloud Computing mean for users?
The first thing it means is that they have no visibility of the physical structure / configuration of the data centre – long gone are the days of wandering over to the rack and re-patching your own connection. But, what visibility do they actually need?
Does it matter if there are separate servers for file & print and for email or whether they are the same physical box?
Does it even matter if the physical servers actually support applications for multiple companies?
There are a host of technologies to manage resources in a data centre – virtual machine (VMware for instance) environments permit a raft of cunning management tricks.
|
None of these techniques are cloud computing, but they are an essential part of it.
Let's move our attention away from an abstract description of a typical corporate data centre to some of the biggest web-based data centres around.
Consider some mythical web-based email service which is capable of handling potentially millions of concurrent sessions, both receiving and sending email along with clients viewing email on-line. There is a requirement for enormous amounts of computing resource to manage these operations – often thousands or hundreds of thousands of virtualised servers running on somewhat fewer physical servers. Here load-balancing is absolutely essential, as is the ability to seamlessly migrate virtual machines away from crashed or damaged physical servers.
So, what makes this cloud computing? The fact that you don't much care where in the world your server resources are located, as long as you can access them from wherever you are.
Really, this means that the Internet is the cloud and we simply choose to represent that the services we are using are there as well (hands up all those that remember the structure diagram that always showed the telecommunications infrastructure as a cloud-like object on a whiteboard).
Coming back to the title of this piece, why does this mean that 'clouds' can be green?
Cloud computing, whatever it is, allows consolidation of back-end resources, allowing greater utilisation of these resources.
|
As owners of such a system, we could buy the biggest and best computers capable of hosting multiple applications and possibly multiple clients on each of them. Clearly there would need to be all kinds of server virtualisation technologies in place to achieve complete separation of the environments, but this is trivially achieved with abstraction products such as the VMware range.
There are a few other problems to solve, such as the ability to move running servers onto different hardware for load-balancing and fault management needs – all of these services are provided out-of-the-box with the majority of the virtualisation products.
In addition, the ability to seamlessly relocate a running server instance to another location as part of a disaster-recovery strategy is also essential – products from Quest Software will handle this need.
So, to round this off, why is the cloud green?
By reducing the requirements for infrastructure both at the client-side and the server-side, clear and obvious energy savings can be made.
|
Emerson Network Power recently offered a new measure of server output: CUPS (Compute Units Per Second) which will lead to a measure of server efficiency when rated against power consumption. In raw terms, CUPS has increased around 7x between 2002 and 2007 based on a combination of both CPU efficiency and raw speed. During the same period, server power requirements increased just 1.8 times.
On this basis, if computing demand in 2007 was the same as in 2002, just 1/8th the energy would be required to achieve it.
This alone would be a major greening of server farms however, when combined with the advantages of systems consolidation, major improvements can be made.
Politically, there is a downside, especially for SMEs. Ask their accountants about the importance of maintaining physical assets. "Dammit – I bought this thing, it's in my server room and by gosh we're going to keep using it!"
However, there are plenty of options (most of which affect the hip pocket) to turn their heads… more in the next instalment.