Machine Learning is nowadays one of the most common expressions in literature, but the fields of application are so vast that it can sometimes be difficult to grasp all the issues involved. In practice, the Machine Learning can be applied to almost all sectors of activity, whether it is the autonomous driving of vehicles in public spaces or learning from the behavior of users of a company information system. While the most compelling results for the general public are found in natural language and vision processing, the application of artificial intelligence to more austere areas provides significant advances and results.
This article will illustrate 2 situations in which automated learning of user behavior will lead to significant gains in productivity and ergonomics, which cannot be achieved by conventional methods.
Increase the user-friendliness of work environments
Today, one of the most powerful and efficient ways to share enterprise IT resources is to virtualize these resources before making them available to users. This publication mechanism allows users to avoid the hardware constraints of desktops, to provide flexibility and mobility for users, and to reduce the IS administration costs. However, the loading time of user applications is a strong element for user convenience, and the best virtualization technologies allow for an average of about 15 seconds to be achieved between the moment the user clicks on the application icon and the moment the application is operational.
In this context, the introduction of components based on artificial intelligence mechanisms takes on its full meaning. Using the capabilities of the Machine Learning machine, it is possible to learn users’ work habits and predict their future behavior.
In fact, automatic learning makes it possible to characterize users’ habits in a probabilistic way and to “construct” a statistical cartography of the use of IS resources and to anticipate their use.
In concrete terms, the AppliDis Booster algorithm makes it possible to predict (with an adjustable probability level, see below) that Ms. Doe will connect every day of the week between 8:30 and 8:45 a. m. to launch the AppX business application; the virtualization solution can therefore preload this AppX application every day of the week at 8:25 a. m. However, since the recovery of an application takes place on average in 1 second, Ms. Doe will have access, in 1 second, to the application she needs to accomplish her mission. This time saving is a major advantage for occupations where mobility is extremely high and the time spent working on each workstation is low compared to connecting kinematics. Artificial intelligence techniques therefore make it possible to provide innovative and more efficient services to users.
Optimize the use of company resources
Predictive loading of applications obviously has a counterpart linked to the consumption of resources necessary to anticipate the need. This additional cost is linked to the forecasting time (generally very low) and to the CPU and memory resources associated with application preloads, since applications will be started earlier than in the absence of predictive operation.
In advanced virtualization solutions that integrate Machine Learning technologies, administrators can balance performance gain and increased resource consumption, for example by differentiating between groups of users with different levels of comfort. The real-time visualization of the impact of application load forecasting on consumed resources allows administrators to optimize these parameters.
Automated learning of user behavior also allows for detailed analysis of the use of IS resources. Based on the accumulated, formatted and analyzed data, Machine Learning can detect patterns that may not have been identified by the IS administrators. These models can be linked to the topology of the infrastructure or to the use of company resources over time. For example, a periodicity of access to them of 8h, 12h or 168h which can lead to slowdowns or blockages. Similarly, the detection of user groups accessing the same set of resources over a given period of time may justify the modification of the physical implementation of associated servers.
Virtualization and Artificial Intelligence, an innovative and powerful duo
Artificial intelligence has become an essential tool for continuously and autonomously analyzing the mass of data produced exponentially by the new uses of IT tools in companies. Thanks to the very large volumes of data and to material technological advances, the precise identification of trends is now accessible: this is the domain of predictive analysis based on data in which the principle is no longer to define rules producing deterministic results, but to learn through experimentation in order to establish correlations. The paradigm shift is therefore fundamental since we are moving from a deterministic model to a probabilistic model.
The main benefit of this approach is that it is now possible to deal with subjects for which there are no rules that are easy to identify and implement. Thanks to its mastery of virtualization and Machine Learning technologies, Systancia provides innovative and high-performance solutions to improve the agility of companies and their ability to deliver access to IT resources in a flexible and efficient way.