Recently attended an “IT Question Time” panel discussion run by A3 Communications on Artificial Intelligence and Machine Learning (ML). The event was really good, got some great speakers in a room to share ideas, thoughts and, (most importantly) actual machine learning experience.
Most of the stories that were shared revolved around business datacentre optimisation, functions optimisation, and new systems that were being trialled. The word ‘Cloud’ was said around 20-30 times which, believe it or not, is actually quite a small amount for these kinds of meetings. However, the word ‘network’ was only paired with cloud once in the entire session.
Like at many future-looking meetings about corporate IT, there is an assumption that the network will grow to just work and will be able handle whatever our needs are in the future. However, people working on my side of the industry know this is definitely not the case.
One of the panellists, David Cumberworth from Virtana (an AIOps and real-time monitoring platform for IT Infrastructure) discussed the reformation of IT at a national consumer retail chain. This company wanted to move all of their systems to the cloud, to aid with the on-site processes at their stores rather than having each store phone home to the corporate datacentre. However, rather than updating or transforming their systems to make them cloud-ready, the company just wanted to move them to the cloud and host them in a cloud datacentre. Needless to say, the old systems were huge, inefficient and clunky, in the way that on-premise systems can be and cloud systems must not be.
Machine learning was used to help map the processes that were in operation, and plan and configure a new, more streamlined cloud-version of the systems that were needed. Now companies might consider that this is unnecessary, and that the cloud is more than capable of handling anything they throw at it. To be fair, if you have the money and want to pay for the increased datacentre needs it probably could.
However, your average branch site, or remote site, is not necessarily going to have the bandwidth to access a huge cloud solution on a regular basis. The solution of throwing money at the problem doesn’t exist if the local core network cannot handle the new level of traffic. Even where bandwidth is available, there are now multiple competing services all vying for your network bandwidth since you are moving everything to the cloud. This is why you need to streamline services and find a way to make corporate systems more data-efficient and bandwidth-efficient as well.
This is a lesson we are seeing as companies make the transition to cloud services. Corporate networks are being put under more and more pressure as cloud adoption skyrockets. This is where machine learning when it is ready for broader adoption can be utilised for the most immediate gain. It can help process and update older systems to make them more data and bandwidth efficient.
So how does this tie back to networking? We are seeing a trend towards smarter networks through SDN and SD WAN. These new ways of viewing the network are providing much more data and visibility at every step, which is already leading to self-healing networks and more efficiency. When this is combined with the intelligence about the applications and how they are designed that is being gained by machine learning fuelled digital transformation, we can see a predictable pattern where both network operations and the systems that rely on them will both grow more optimised as they learn from each other.
We are getting into realms of optimism now, but it is easy to see how intelligent machine learning making systems more efficient combined with reactive and intelligent network management can not only reduce network burdens but also allow them to perform new feats that we are only now beginning to understand.