There’s another phrase that has recently crept into more common parlance amongst IT vendors -  `employee-owned IT’. This is a concept that has the potential to be one of the better two-edged swords the business has ever come up with, but it is also a model that is the most likely to move everyone, ever faster, towards using remote infrastructure models based around The Cloud. 

From the point of view of any business it is something to be welcomed, for it moves the cost of end-user client systems from the business and onto the employees. The possible downside, of course, is that it may also end up as a tax, imposed by, and paid to, the employer, for the privilege of having a job. In the end, employers will need to pay bigger salaries as compensation, particularly for PAYE staff. And they will need to claim this stuff against tax, of course. 

While the coalface users might like the option of choosing their own client systems – be it on price or fashion consciousness, this opens up all sorts of problems for the employers. These things need to be standardised and controlled. Allowing staff to use what they want as a client can only cause endless grief for the IT department unless very strict controls are applied to what is bought. This can mean buying clients from the `company store’, which can look even more like a tax on having a job. 

But worst of all the employee-owned IT model has real security problems attached to it, not least being the standard ones of users either losing their clients, or having them stolen – and for every lost MOD laptop that makes the news, there a bound to be many more from private companies that hushed up (and probably more damaging). Superglueing the thing to users’ hands only shows the bad guys that this is a laptop worth stealing. 

The only sensible way of exploiting employee-owned IT is to not let them near the data. And the only way to do that is with portable, personal, thin clients that are only portals through to the corporate data back at the datacentre. But the data never leaves there and is never stored on the device itself. Once the connection is severed there is nothing worth stealing except a bunch of bits. 

And there are several ways the connection between portal and datacentre can be quickly severed in an emergency – a large red panic button, for example, or perhaps a thin, weak, chord to a wrist strap worn when the device is in use – and it won’t work without the strap fitted. Pulling the client from the user when in use breaks the chord, and the connection. 

Meanwhile the data is safe, and that is the primary role of any infrastructure. It also means that, even if the business has and onsite datacentre, it is effectively running it as an end-to-end, integrated service delivery utility. It will be the only way the employee-owned IT model can work securely. From there, it will become a much shorter step to decide just how much of an existing onsite infrastructure is actually needed next time there is a major upgrade and how much can be moved to a Cloud-based remote infrastructure provided by specialists.

The question is: do we need some new metrics and points of reference when it comes to the relationship between IT, its users, and its carbon foot-print? 

It is estimated that IT already has a carbon footprint as big as the airline industry – 2% of all carbon emissions. Somehow, that already seems a little low and not that unreasonable. After all, though many would argue about the aggravations caused by IT – Little Britain’s archetype: `the computer says “no”’ now being part of urban folklore  - its benefits-per-Kg of carbon is quite possibly better for most people than the aircraft – and most certainly will be in the future. 

More importantly, however, its carbon footprint is growing fast, while so far managing to avoid the public opprobrium now meted out to the airline business. So there is still time for the IT industry – vendors and users – to grasp the nettle of cutting energy consumption and reducing carbon foot-prints before the rest of the world notices the problem. 

Most of the big systems vendors are already trundling down this track, but there still seems a reluctance amongst users to readily jump on the bandwagon they are collectively pushing. Which is why the question arises about the need for new metrics that make the point to users more effectively. 

Sun Microsystems certainly seems to be thinking along this line, and is keen to use itself as an example. It has re-engineered its own technology and business procedures – for example, 40% of staff now work from home – and in the process has saved an estimated 29,000 tons of CO₂. This is important, but in IT usage terms perhaps a rather meaningless metric. Sun’s analogy - the equivalent of 500 planes flying round the world 10 times – doesn’t help much either: I suspect that equates to about an hour’s worth of global air travel (approximately 120,000 1 hour flights).  

A more directly obvious metric is the T5000 servers, which 10 years ago would need 128 servers in three racks, consuming 120 kW. Now it is one box using 1kW. More relevant still for many large businesses is the Sun Ray thin client, which has 5% of the energy use of a PC. With no moving parts it also has a lifecycle of 20 years or so – and all the intelligence is back at the server. And server technology is allowing significant growth in power and performance coupled to a reduction in footprint, cooling requirements and energy consumption. This combination can actually deliver growth in business service with a reduction in both energy costs and TCO, for there are also lower maintenance costs. 

For example, if a thin client uses only 5% of a PC’s energy, a large company with several thousand of them can afford to re-invest in newer, faster servers every five years, with each new server running more thin clients than the server it replaces? If the lifecycle of the thin clients is 20 years, that is four iterations of server development, each one capable of handling more client workloads than the last. Now the energy running cost per server is more than the cost of buying them, this is a model that allows users the flexibility to choose between a future of fewer servers delivery the same workload and the same number offering real growth potential.  

A specific example comes from Sun lifecycle management specialist partner, Bell Microsystems, and its customer, Rightmove. They replaced old servers with T1000 units which now deliver 500K page impressions/month with a energy saving of 92%, a similar saving on the physical footprint, and a saving of 9600 kgs of carbon/month. All must be useful at a time of a downturn in the housing market. 

What about road warriors and their need for laptops? Well Sun staff already come equipped with smart cards for Sun Ray clients, which is a model that can – with sufficient security and engineering – be extended to every hotel room at the very least. So there is a need for some new metrics for both users and vendors to work with. The ultimate figure, of course, is the overall cost of delivering relevant processes and services that are of definable value to the `user’ (business, organisation or individual) set against the cost of constructing the systems in the first place (ideally including the social costs of the impact on the raw materials excavation and processing), the costs of using and the increasingly important costs of disposal and savings of re-use. 

The need to change from bangs-per-buck to bangs-per-carbon/Kg. Define the relationship between CPU cycles/sec and what is being delivered. This is a lifecycle management issue every CIO is now facing - what application is used to meet a business requirement, how well is it written, how well the infrastructure is architected and implemented.

I love analogies, and two immediately come to mind following Bill Gates’ keynote at the recent Microsoft TechEd event. How about: Microsoft’s move into mega-datacentres and Software as a Service is like using the latest race car aerodynamics theories to design a horse and cart. It should go fast in theory, except that it is still actually pulled by horses – lots of them, to be sure, but still horses. Or, it’s like deciding that the best way to get to the moon is to start by building a very, very long bridge.   Gates’ Keynote showed that he has discovered The Cloud, but it seems he is still viewing it as just another way for Microsoft to deliver `applications’ rather than as a platform for `services’. This poses two alternative problems, one for users and one for Microsoft itself.  

If Microsoft succeeds with promoting the idea of The Cloud as an application delivery vehicle, and it certainly has the marketing muscle to achieve it, users could well find themselves disadvantaged. If they feel that they are still working with traditional applications in the traditional manner, it may well give them a great feeling of comfort. But it is quite likely that they will soon find themselves increasingly disadvantaged against competitors that do adopt a service-based approach. They are likely to gain greater operational flexibility and agility in the way they can serve their respective markets, meet changing customer demand, and adapt their business processes in order to manage those changes. 

The other problem if Microsoft succeeds is that considerable confusion may exist amongst users. The whole issue of moving users from a pay-up-front permanent licence acquisition business model to an annuity model is one of the main stumbling blocks of the shift to any service-based, per-per-use model, particularly from the vendors’ point of view. If Microsoft tries to sell the idea on the basis of users still `having their applications’ it could end up with a good deal of egg on its face from the mixed messages.  

A side issue of this is that Microsoft is joining the race to build vast datacentres, all obviously based upon multicore x86 architecture processors. But this itself poses a number of problems, not least being that the levels and volumes of service that will be required will push heavily towards the need for complex, parallel processing architectures. There are some doubts here as to how effective large, Windows-based multicore x86 devices can be in such an environment. It has to be noted that IBM’s latest supercomputer now makes extensive use of Cell architecture processors to break the PetaFlop barrier.  

Windows as an operating system has just about reached the limit of its capabilities with 8-core processors, while Microsoft’s recent track record in this area -
Vista showing that it seems to have started to miss the sweet spot for business users – suggests that this trend may become more prominent as datacentres become huge and hugely oriented towards service-delivery. There has to be some doubt as to whether the company now has the right background for building systems capable of running a Very Large Datacentre delivering billions of dynamically assembled, loosely coupled code components as millions of individual, short-lived service entities.