Search papers, labs, and topics across Lattice.
This paper introduces SLO-Scaler, an innovative autoscaling framework that leverages a Bayesian LSTM model to predict request rates, tail latency, and SLO violation probabilities for microservices. By incorporating uncertainty awareness and dependency graph analysis, SLO-Scaler effectively localizes bottlenecks and avoids unnecessary scaling across the entire service chain. The evaluation on the DeathStarBench Social Network benchmark demonstrates that SLO-Scaler significantly reduces SLO violation rates by 29-56% and scaling event frequency by 38-59%, while also maintaining lower tail latency compared to existing methods.
SLO-Scaler cuts SLO violation rates by up to 56% while reducing scaling events by nearly 60% through uncertainty-aware predictions and targeted scaling.
Autoscaling microservice-based applications to satisfy Service Level Objectives (SLOs) remains challenging due to bursty workloads, cascading latency across service dependencies, and cold-start overhead. Existing approaches such as the Kubernetes Horizontal Pod Autoscaler (HPA) rely on threshold-based CPU or memory metrics, which react too slowly to traffic spikes. Recent predictive methods improve responsiveness but generate point forecasts that ignore prediction uncertainty, leading to over-provisioning or oscillatory scaling. We propose SLO-Scaler, an uncertainty-aware autoscaling framework that predicts short-horizon request rates, tail latency, and SLO violation probability using a Bayesian LSTM model. SLO-Scaler integrates confidence-interval-based scaling decisions with a dependency graph analysis module that localizes bottleneck services, avoiding unnecessary whole-chain scaling. We evaluate SLO-Scaler on the DeathStarBench Social Network benchmark deployed on Kubernetes under periodic, bursty, and long-tail traffic patterns. Under bursty traffic, SLO-Scaler reduces the SLO violation rate by 29-56%, lowers the average replica count by 18-33%, and decreases scaling event frequency by 38-59% compared with the baselines, while achieving lower tail latency.
