International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 05 | May 2025
p-ISSN: 2395-0072
www.irjet.net
Using Recommendation Techniques to Reduce Cloud Monitoring Fatigue Priyanka Mishra1, Siddharth Shroff 2 1Senior Product Manager, Microsoft, California, USA 2 Principal Product Manager, Amazon, California, USA
---------------------------------------------------------------------***--------------------------------------------------------------------recommendation algorithms sift through millions of Abstract - Modern cloud ecosystems produce enormous
options to highlight the few items most relevant to each person. They excel at filtering noise and prioritizing what matters to the end-user.
volumes of telemetry-metrics, logs, and traces that commonly inundate engineering teams with noisy or lowpriority alerts. In this paper, we discuss the use of recommender system concepts in cloud observability for filtering and prioritizing operational data selectively. Using techniques such as collaborative filtering and pattern finding, observability platforms can bring the most pertinent signals to the forefront, mitigate alert fatigue, and accelerate root cause analysis. The article defines the crossroads of monitoring and AI recommendation, proposing a practical strategy for technical leaders to build more responsive, human-centric, focused, and efficient operations workflows.
Key
This raises an intriguing question: can we apply the same recommendation techniques to filter and prioritize observability data, thereby reducing cloud monitoring fatigue? This paper explores applying these principles from recommender systems to enhance cloud observability, aiming to reduce monitoring fatigue and sharpen operational focus for technical leaders. 1.1 What is Cloud Observability? Cloud observability is the practice of understanding the internal state of systems by examining the telemetry (metrics, logs, traces, etc.) they produce. Unlike traditional monitoring which checks a few preset metrics, observability provides a comprehensive, proactive view into system behavior. It allows engineers to detect, diagnose, and understand issues efficiently. Observability is investigative and helps teams move beyond “Is the system up?” to “Why is the system behaving this way, and what does that mean for customers?”
Words:
Observability, Cloud Computing, Recommender System, Recommender Systems, Monitoring Automation, AI-driven Operations, Noise Reduction in Monitoring, Anomaly Detection, AIOps
1.INTRODUCTION Cloud computing enables enormous scale with incredible flexibility, but it has also led to a deluge of monitoring data and alerts. However, these large-scale systems are complex and have many moving parts; even a small issue can quickly snowball into a bigger problem. Modern observability tools generate a firehose of telemetry, often triggering excessive alerts – many low-priority or false alarms, resulting in alert fatigue. This happens when engineers, bombarded with notifications, start tuning them out, much like ignoring car alarms that cry wolf too often. Consequently, important warnings can get missed [4]. Imagine a search system that starts serving completely irrelevant results because a data pipeline silently broke hours ago, but the alert for it was buried among dozens of less critical ones. As a result, incident response slows down while customers have a poor experience, and the team scrambles under pressure. Over time, this takes a toll, burnout becomes real, and trust in the alerting system erodes.
There are three key inputs to a powerful observability system- (1) Metrics are numerical measurements (e.g. CPU utilization, request latency, error rates) that show how a system is performing over time. They’re great for spotting trends, setting alerts, and tracking service health. (2) Logs are detailed notes, timestamped records of events (e.g. an application error, warnings, or a user action) that help pinpoint exactly what happened, when, and why. (3) Traces track and reconstruct the end-to-end flow of a single request or transaction through distributed components, which is invaluable for debugging in microservice architectures. They help pinpoint where slowdowns or failures happen in complex, distributed applications. Together, these telemetry streams give a rich, highcardinality view of a cloud system’s operation so teams can quickly detect issues, understand root causes, and improve performance. Modern cloud observability
Meanwhile, in a completely different domain, recommender systems have evolved to tackle information overload for users. From e-commerce to streaming media,
© 2025, IRJET
|
Impact Factor value: 8.315
|
ISO 9001:2008 Certified Journal
|
Page 848