Optimizing Service Level Agreement Tier Selection in Online Services Through Legacy Lifecycle Profile and Support Analysis: A Quantitative Approach

This study introduces a novel approach to optimal Service Level Agreement (SLA) tier selection in online services by incorporating client-side obsolescence factors into effective SLA planning. We analyze a comprehensive dataset of 600 million records collected over four years, focusing on the lifecy...

Full description

Saved in:
Bibliographic Details
Main Authors: Geza Lucz, Bertalan Forstner
Format: Article
Language:English
Published: MDPI AG 2025-05-01
Series:Mathematics
Subjects:
Online Access:https://www.mdpi.com/2227-7390/13/11/1743
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This study introduces a novel approach to optimal Service Level Agreement (SLA) tier selection in online services by incorporating client-side obsolescence factors into effective SLA planning. We analyze a comprehensive dataset of 600 million records collected over four years, focusing on the lifecycle patterns of browsers published into the iPhone and Samsung ecosystems. Using Gaussian Process Regression with a Matérn kernel and exponential decay models, we model browser version adoption and decline rates, accounting for data sparsity and noise. Our methodology includes a centroid-based filtering technique and a quadratic decay term to mitigate bot-related anomalies. Results indicate distinct browser delivery refresh cycles for both ecosystems, with iPhone browsers showing peaks at 22 and 42 days, while Samsung devices exhibit peaks at 44 and 70 days. We quantify the support duration required to achieve various SLA tiers as follows: for 99.9% coverage, iPhone and Samsung browsers require 254 and 255 days of support, respectively; for 99.99%, 360 and 556 days; and for 99.999%, 471 and 672 days. These findings enable more accurate and effective SLA calculations, facilitating cost-efficient service planning considering the full service delivery and consumption pipeline. Our approach provides a data-driven framework for balancing aggressive upgrade requirements against generous legacy support, optimizing both security and performance within given cost boundaries.
ISSN:2227-7390