Optimising web interface performance using Amdahl`s law

Abstract

The article discussed the construction and application of generalised mathematical models based on Amdahl’s law to determine the maximum possible acceleration of web interfaces, taking into account their key features. An extension of the classical approach was proposed by including asynchronous processes, multi-level caching mechanisms, and dynamic resource loading methods in the model, which allows for a more accurate assessment of the cumulative impact of various optimisations on performance. In particular, the feasibility of taking into account asynchronous data exchange was justified, which allows processing requests in parallel and avoiding blockages in the process of updating content. A formula has been developed that takes into account the effectiveness of client and server caches and provides a quantitative assessment of the reduction in response time when reusing already loaded data. Particular attention was focused on step-by-step content retrieval techniques, where the initial page load was minimised by deferring the addition of individual scripts, images or styles, which speeds up the initial display of important content and makes the interface more responsive to user actions. In addition, the impact of a comprehensive combination of optimisation strategies on web interface performance was considered, and a corresponding generalised model was proposed, which uses an interdependence coefficient to determine the extent to which one optimisation enhances or, conversely, negates the effect of another. This makes it possible to predict the total performance gain and compare the cost of implementing several solutions with the potential time savings. The proposed formalised approach can serve as a basis for creating automated tools for evaluating web interface performance, integrated into the development process. Testing the model in three practical scenarios – partial rendering with API caching, JavaScript minification with a content delivery network (CDN), and code splitting with server-side caching – yielded performance gains of 1.87×, 1.55 ×, and 1.64 ×, respectively, which was fully consistent with theoretical predictions. The data obtained confirmed the ability of the R interdependence coefficient to accurately reflect the synergy or overlap of optimisation effects and makes the model suitable for pre-selecting the most effective acceleration strategies at the CI/CD audit stage.