Trial and Error

As well as selection of over complicated technology products there are other factors that may well contribute to the slow deployment of technology. As we have aspired to IT professionalism we have also potentially lost some of the factors that made us agile.


We have decided that we should get everything perfect first time, after all we specialise in this stuff don’t we? This is reinforced by the way we are treated by our businesses if even the smallest thing goes wrong. As a breed nothing is every good enough for an engineer in whatever discipline they work. We can always improve something and will do so if we are allowed to, even if that means starting again from scratch and missing deadlines. Add to this the threat of bite back should anything fail and projects take too long as we try harder and harder to make them perfect.

The evolutionary process that has lead to our own existence on this planet and then to all of our achievements makes use of the trial and error process. You try a few things discard the failures then keep the successes and try a few more things based on them. To improve our delivery timescale we need to be able to work in ways that allow application of trial and error.

In software development we have methodologies such as agile that allow us to use this process quite effectively. I have seen a number of significant successful software development projects use this to great effect. In infrastructure though this does not seem to be the case as large investments in equipment that don’t do what was expected is never going to be popular. Once again though there is hope on the horizon in the form of Cloud compute.

Spinning up a computer in the cloud to try out your new service costs you only the computing power that you use. If the service does not work then just turn it off again and stop paying, suddenly you are free to take risks. This potentially allows you to become much more agile in your approach to the creation of infrastructure. Not only that but successful trials can be scaled up within the same service into a live system, or depending on the services you are creating can be moved onto on premise equipment.

Of course software is a factor in here that currently confuses the picture. Whatever software you are putting on the Cloud compute infrastructure will need to be licensed correctly. This may well be a driver that helps certain open source products to proliferate as their licensing may better suit this agile environment, until the large software vendors work that out at least.