Gespeichert in:
| 1. Verfasser: | |
|---|---|
| Format: | Recurso digital |
| Sprache: | |
| Veröffentlicht: |
Zenodo
2026
|
| Online-Zugang: | https://doi.org/10.5281/zenodo.19335510 |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Inhaltsangabe:
- <p><span>Software is changing at an astonishing pace, and because so many teams are using DevOps, the routes for getting changes out (CI/CD pipelines) have become much, much more complicated [1], [3]. Modern DevOps is all about doing things automatically and getting updates live quickly, but the ways we watch over things are still mostly about responding to problems as they happen [5], [6]. That means we aren't good at spotting something going wrong before it does, guessing when things will break, or getting the very best out of our systems [16], [20]. This problem is particularly obvious in cloud-native and distributed systems - where keeping things running reliably and at a good speed is essential but tricky with old-fashioned monitoring [6], [26].</span></p> <p><span>This work presents a smart, data-focused way of monitoring that will improve how DevOps works by bringing together in-depth observability with analytics using machine learning [5], [18]. It’s a single design which pulls in data as it happens from the CI/CD pipelines, the infrastructure itself, logs and distributed traces, and then uses smart methods to find anything odd, predict when failures will happen, and automatically feed that information back into the pipeline [15], [20]. Drawing on ideas from Site Reliability Engineering (SRE) and AIOps, the idea is to change monitoring from a tool you use after something has gone wrong to something that actively makes your DevOps process better [6], [18].</span></p> <p><span>The way we did this was to create and build a smart monitoring system which works with popular DevOps tools and cloud-native platforms like Kubernetes and Docker [23], [24]. This lets us gather and study data that will grow with your system, even as it’s spread across lots of places. We use machine learning to look at how things have gone and how they are now, to find repeating patterns, predict when the system will fail, and suggest ways to improve [16], [20]. We measure how well it does using important stats like how often we get updates out, how long it takes to get back to normal after an issue (MTTR), how long the system is up and running, and how efficiently the pipeline is working [3], [28].</span></p> <p><span>The experiments show that plugging smart monitoring into CI/CD pipelines really does make the system more dependable, gets us responding to incidents far quicker and makes everything run more smoothly [3], [26]. The results show how useful AI-powered observability can be for making smart decisions in advance and letting systems adjust to their environment in complex software worlds [18], [29]. Essentially, this research gives us a way to make DevOps better that can expand easily, and provides the groundwork for building software delivery systems that can fix themselves.</span></p>