21.08.2019 l by Meinrad Happacher
Continuation of the Article of Part 2.
The cloud scales itself as needed
There is another way: The Cologne-based cloud provider gridscale, for example, does not define the cloud from the supply side, but from the user side. Each instance has a fixed price, which is calculated according to usage. This is transparent and easy to handle. If required, instances can be quickly switched off via the API and then immediately no longer incur any costs.
But cloud usability can mean even more than simple handling: a platform as a service, for example, which adapts itself elastically to the business as a fully automated cloud operation, at best without user intervention. Such an intelligent algorithm enables a number of data-based cloud services. Based on the experience gained, the system can forecast how workloads will develop. Additional capacity is automatically available when it is needed - autoscaling. It is also possible, for example, to migrate workloads live. They can be redistributed, for example, because maintenance work has to be carried out on the server or because they need to be relocated to a cheaper resource. There is no doubt that attempts to break into the IT infrastructure are also among the anomalies - and these can also be detected at an early stage using certain parameters, such as specific patterns in network traffic.
Artificial intelligence is thus finding its way into the data center. A machine learning algorithm monitors all metrics of the cloud environment, such as CPU utilization and temperature, the number of I/O accesses, latency times and much more. In a longer process, the algorithm learns from practical events what anomalies can be measured against normal operation and what measures to take. The goal behind machine learning in the cloud becomes clear: the user defines which parameters the cloud should meet in terms of performance, costs, availability and security. The intelligent algorithm implements these parameters and continually adapts the cloud accordingly.
The whole original article in german can be found here.