Autonomous Runtime Management of Cloud Performance Anomalies Using End-to-End Service Topology
Performance anomalies are a norm rather than an exception in cloud application services. Ensuring Service Level Agreements (SLAs) during such anomalies poses a significant challenge for cloud service providers due to the large number of components involved and their complex interactions. The most effective operational strategy for managing these anomalies is to strengthen application service resilience by implementing autonomous capabilities for detection, diagnosis, and remediation. Crucially, these capabilities must operate at the level of the application service itself, rather than focusing on individual layers of the cloud stack. To address this, the paper introduces Topology-Aware Autonomous Recovery (TA-REC), a proof-of-concept framework that leverages a closed-loop MAPE-K system for automated performance remediation. TA-REC is an autonomic manager that detects deviations in performance SLOs as anomalies, localizes their root causes, identifies the affected end-to-end service components, and remediates the anomalies using topology-aware recovery strategies. To achieve this, TA-REC leverages state metrics from end-to-end service components to detect performance anomalies and activity metrics to guide the selection of recovery actions. Evaluations across various fault scenarios that reflect real-world conditions show that TA-REC consistently identifies the optimal recovery action for each scenario, taking into account both Recovery Percentage and Recovery Time. The results indicate that, during the evaluation window, the Recovery Percentage ranges from 0.57 to 1.29.
Tue 30 JunDisplayed time zone: Brussels, Copenhagen, Madrid, Paris change
11:00 - 12:30 | |||
11:00 60mKeynote | Between static and run-time verification of message-passing systems with behavioural types VORTEX | ||
12:00 25mTalk | Autonomous Runtime Management of Cloud Performance Anomalies Using End-to-End Service Topology VORTEX Dhanya R Mathews India Indian Institute of Science, Bangalore, Jagarlamudi Lakshmi Indian Institute of Science, Bangalore, | ||