From future to present: smart data centres
In the first place, it should be pointed out energy costs increase annually by around 10% on average on a global scale, being Spain one of the countries which is above it. To quantify this fact, data centres from just the United States consume an average of 90 billion kWh per year, while the rest around the world use around 416 terawatts of electricity in the same period of time. Seeing the excessive consumption of electricity, there are implementations of artificial intelligence to save energy by avoiding overheating of the equipment. This saving is possible through a prediction of the future workload and a load distribution system, reducing costs and significantly improving the efficiency of the system. In fact, hyperscalars are implementing this technology given the improvements in performance, device monitoring, optimization in processing times and speed in reducing risk factors. In a near future, we may see how these centres will be able to automate most of their operations given because, although workloads will continue to evolve with the increased amount of data moving around, AI will also do it too at the same time allowing to find new uses for it.
Secondly, another area in which AI is being decisive, especially in the data centre environment, is in everything related to the reduction and resolution of accidents. On the one hand, it is becoming more and more common to find data centres capable of self-management through tools such as deep learning to predict failures ahead of time. It should be noted that behind the aforementioned deep learning the data is found, which goes far beyond what can be extracted from records and metrics of traditional hardware platforms, since companies do not store their information only in the data centre but also in a hybrid way making use of the cloud. On the other hand, these are also usually accompanied by recommendation systems based on machine learning that allow any problem to be located and tackled in advance. The learning carried out must always be available where and when it is needed, otherwise the costs would rise and the efficiency would be affected, decreasing accordingly. As can be seen, intelligent monitoring, data storage and decision-making based on the latter once again make an appearance, thus forming a competitively robust accident prevention system.
Third and last, another feature that is gaining momentum is logical security within data processing centres. Currently, there are systems capable of introducing, from time to time, data collection probes that allow to correlate the different stored logs making possible to automatically detect and block difficult to prevent attacks, being the Advanced Persistent Threats (APT) a clear example. Thus, this technology is able to analyse and identify normal behaviour in networks, making it easier to detect a situation of cyber risks based on anomalies and deviations in the infrastructure.
In order to prepare the project to be carried out in the subject of "Network management and planning", further research will be carried out on the inclusion of all these possibilities that AI can offer, assessing its viability in the environment in which our case is developed and you will be kept informed about the progress made in subsequent posts. Thank you very much for your time spent reading us!
Arturo Moseguí and Enric Sasselli