A couple of emails floated into my inbox recently which brought home to me about how long the journey is going to be for many companies as they try to move to a service oriented delivery for IT. I think many are going to be flailing around for some years to come as they try to make sense of ‘new’ paradigms; not just the IT functions but this impacts beyond this.
The technological changes are important but actually, much could be achieved without changing technologies massively. All that is required is mindset change.
Pretty much all traditional IT is delivered based on a presumption based delivery model; everything is procured and provisioned based on presumption.
A project will look at its requirements and talk to the IT delivery teams; both teams often make the presumption that both sides know what they are talking about and a number of presumptions are made about the infrastructure which is required. An infrastructure is procured and provisioned and this becomes often a substantial part of the project costs; it is also something which is set in stone and cannot change.
I don’t know about you but if look at the accuracy of these presumptions; I suspect you will find massive over-provisioning and hence the cost of many projects are overstated. Or sometimes, it is the other way round but examining most IT estates (even those heavily virtualised) there is still lots of spare capacity.
However, you will find that once the project funding business unit has been allocated the infrastructure; they are loath to let it go. Why should we let the other guy get his project cheap? And once a project is closed, it is often extremely hard to retrospectively return money to it.
Of course, this is nonsense and it is all money which is leaving the Business but business units are often parochial and do not take the wider picture into account. This is even more true when costs are being looked at, you don’t want to let the other guy look more efficient by letting them take advantage of your profligacy. It is politically more astute to ensure that everyone is over-provisioning and ensuring that everyone is equally inefficient!
In IT, we make this even easier by allowing an almost too transparent view into our provisioning practises. Rate-cards for individual infrastructure components may seem like a great idea but it encourages all kinds of bad practise.
‘My application is really important, it must sit on Tier 1` has often lead to a Tier 1 deployment fair in excess of what is really required. However if you are caught moving a workload to a lesser tier, all kinds of hell can break out; we’d paid for that tier and we are jolly well going to use it.
‘My budget is a little tight, perhaps I can get away with it sitting on a lower tier or not actually provision enough disk’; I’ve seen this happen on the grounds that by the time the application is live and the project closed; it becomes an IT Support problem. The project team has moved on and its not their problem.
The presumption model is broken and leads to dissatisfaction both in the IT teams and the Business teams. In fact it is probably a major factor in the overwhelming view that IT is too expensive.
The consumption model is what we need to move to but this does mean some fundamental changes to thinking about IT by Business Leaders and IT Leaders. If you want to retain a private IT infrastructure and many do; you almost have to take a ‘build it and they will come approach’; the Service Provider competitor already does this, their model is based entirely on this.
You need to think about your IT department as a Business; however, you have an advantage over the external competitor or at least you should.
- You should know your Holding company’s Business.
- You only have to break even and cover your costs, you do not need to make a profit and any profit you do make should be ploughed straight back into your business. This could be in the form of R&D to make yourself more efficient and effective or it could be on infrastructure enhancement but you do not have to return anything to your shareholders apart from better service.
- You should have no conflicting service demands; there should be no suspicion that another company is getting a better deal or better service. You can focus! You can be transparent.
When I talk about transparency, you should beware of component level rate cards; you should have service rate cards based on units consumed; not units presumed to be allocated. In order to do this, you will need a dynamic infrastructure that will grow to service the whole. It would be nice if the infrastructure could shrink with reduced demand but realistically that will be harder. However, many vendors are now savvy to this and can provision burst capacity with a usage-based model but beware of the small print.
There might be ways of using redundant capacity such as DR and development capacity to service peaks but this needs to be approached with caution.
And there is the Holy Grail of public Cloud-Bursting but currently most sensible experts believe that this is currently not really viable except for the most trivial workloads.
If you have a really bursty workload, this might be a case when you do negotiate with the Business for some over-provisioning or pre-emptable workloads. Or you could consider that this is an appropriate workload for the Public Cloud, let the Service Provider take the investment risk in this case.
But stop basing IT on presumption and focus on consumption.
