<p> Cloudified mobile networks, such as 5G, are expected to deliver a multitude of services to several slices in parallel, while having reduced capital and operating expenses. The 5G mobile systems, therefore, need to ensure that the SLAs of customized end-to-end sliced services are met. This requires monitoring the resource usage and characteristics of data flows at the virtualized network components and interfaces of its cloud mobile network, as well as tracking the performance at its radio interfaces and UEs. A centralised monitoring architecture can not scale to support millions of UEs though. This paper, proposes a distributed telemetry framework in which UEs act as early warning sensors. Upon flagging an anomaly, the cloudified mobile network activates a machine learning model to attribute the cause of the anomaly. We employ active, passive and in-band telemetry in our monitoring framework<br> and achieve an impressive performance of 85% F1 score in detecting anomalies caused by different bottlenecks, and an overall 89% F1 score in attributing these bottlenecks. Our distributed framework achieves almost same bottleneck attribution accuracy to that of acentralized monitoring system but with no overhead of transmitting UE-based telemetry data to the centralized controller. <br>  </p>

<p> Cloudified mobile networks, such as 5G, are expected to deliver a multitude of services to several slices in parallel, while having reduced capital and operating expenses. The 5G mobile systems, therefore, need to ensure that the SLAs of customized end-to-end sliced services are met. This requires monitoring the resource usage and characteristics of data flows at the virtualized network components and interfaces of its cloud mobile network, as well as tracking the performance at its radio interfaces and UEs. A centralised monitoring architecture can not scale to support millions of UEs though. This paper, proposes a distributed telemetry framework in which UEs act as early warning sensors. Upon flagging an anomaly, the cloudified mobile network activates a machine learning model to attribute the cause of the anomaly. We employ active, passive and in-band telemetry in our monitoring framework<br> and achieve an impressive performance of 85% F1 score in detecting anomalies caused by different bottlenecks, and an overall 89% F1 score in attributing these bottlenecks. Our distributed framework achieves almost same bottleneck attribution accuracy to that of acentralized monitoring system but with no overhead of transmitting UE-based telemetry data to the centralized controller. <br>  </p>