Analyzing the Efficiency of Software Distribution Systems in AIoT Environments

In the past few decades, the capabilities of hardware platforms designed for the development of specialized devices (such as mobile phones or embedded computers) have advanced tremendously. One significant consequence of this evolution is the opening up of possibilities for using standard software (e.g., the Linux operating system, standard libraries, etc.) on these platforms. The latest phase of this evolution is represented by the emergence of AIoT (Artificial Intelligence of Things) devices, specifically IoT devices that have sufficient hardware resources to execute complex and demanding artificial intelligence algorithms locally. In (A)IoT environments, especially in cases where devices are battery-powered and their autonomy is crucial, performance and algorithm optimization can often be more significant criteria than result accuracy and precision. Until recently, there has not been much focus on optimizing the distribution of software solutions. However, due to several factors, such as a shift in software development philosophy toward more frequent and smaller changes, limited resources in IoT environments, the need for automated updates, increasingly bulky software packages, etc., the question of efficient software distribution and related updates is becoming increasingly important. In this chapter, we provide an overview of standard solutions and approaches for software package distribution and management, as well as an analysis of their efficiency. We consider classical package management systems, package distribution systems for specific programming languages, and hybrid systems of this kind. The analysis primarily focuses on IoT systems characterized by the need for efficient utilization of local resources and communication channels.