"ACROSS - SLA Monitoring Telemetry Dataset "

"This dataset includes compute telemetry data collected from a Server-Client setup deployed across a 5G network in a period of 3.5 days in the context of Test Case 4.6 (TC4.6) for the ACROSS project.The dataset is used to train forecasting models that continuously monitor a system's compute telemetry data and ensure that Service Level Agreements (SLAs) issued by the system's operators are maintained under different system operating conditions. These include SLAs related to CPU usage, memory allocation and bandwidth. When a potential SLA violation is detected, alerts are passed to other system components, such as the Automation Event Handler, which triggers the Policy Executor to evaluate the event and determine whether adaptation is required.The setup includes a Streaming Server that streams a video consumed by a Streaming Client. Both the client and the server are deployed in dedicated Kubernetes clusters. Given the above setup, the collected telemetry data includes multiple univariate measurements from four different points of view:The Streaming Server, deployed as a container in a Kubernetes pod.The Streaming Client, deployed as a container in a Kubernetes pod.The Server K8-Cluster, that is the K8 cluster that contains the Server container.The Client K8-Cluster, that is the K8 cluster that contains the Client container.The measurements were taken from Prometheus endpoints, in 1-second intervals, collecting telemetry data from the services and the K8 clusters with a scrape-period of 30 seconds. For each measured instance, a total of 288,435 values were collected."

本数据集包含ACROSS项目测试用例4.6（TC4.6）场景下，于3.5天周期内，在跨5G网络部署的服务器-客户端架构中采集的计算遥测数据。本数据集用于训练预测模型，以持续监控系统的计算遥测数据，确保系统运维方所签署的服务水平协议（Service Level Agreement, SLA）在不同系统运行工况下均得到严格遵守，相关协议涵盖CPU使用率、内存分配与带宽等方面。当检测到潜在SLA违规行为时，系统将向自动化事件处理程序（Automation Event Handler）等其他组件发送告警，该处理程序会触发策略执行器（Policy Executor）对事件进行评估，以确定是否需要执行系统适配操作。本次部署架构包含一台流媒体服务器（Streaming Server），其向流媒体客户端（Streaming Client）推送视频流；客户端与服务器均部署于专属Kubernetes（K8s）集群中。基于上述架构，本次采集的遥测数据包含四类不同视角的多组单变量测量项，分别为部署于Kubernetes Pod容器中的流媒体服务器、部署于Kubernetes Pod容器中的流媒体客户端、承载服务器容器的K8集群（Server K8-Cluster），以及承载客户端容器的K8集群（Client K8-Cluster）。上述测量数据均取自普罗米修斯（Prometheus）监控端点，以1秒为采集间隔，并以30秒为抓取周期从各服务及K8集群中采集遥测数据。每个被测量实例共采集到288,435条有效数据值。