ADASCALE: An Adaptive Scaling and Placement Framework for Microservices Under Dynamics

A MAPE-loop approach that jointly scales and places microservice replicas to meet SLOs under non-stationary traffic, mixed root operations, and heterogeneous inter-node latency in cloud-edge environments.

Ming Chen; Muhammed Tawfiqul Islam; Maria Rodriguez Read; Rajkumar Buyya· 2026· DOI 10.48550/arXiv.2607.15681

The core problem

Microservice applications are increasingly deployed across cloud-edge environments, where heterogeneous nodes and time-varying inter-node delays amplify the impact of placement decisions. At the same time, these applications face non-stationary traffic, shifts in the mix of root request operations that exercise different call graphs, and heterogeneous communication modes that determine how network latency and queuing propagate to end-to-end (E2E) performance. Existing autoscalers and network-aware schedulers typically handle only a subset of these dynamics, leading to either compute bottlenecks or inflated cross-node latency and thus SLO violations. The authors propose ADASCALE, an adaptive framework that jointly scales and places microservice replicas under such multi-dimensional dynamics. The core problem addressed is the joint optimization of replica count and placement to satisfy SLOs while minimizing latency and maximizing throughput in dynamic cloud-edge settings.

ADASCALE is evaluated on a cloud-edge Kubernetes cluster using the DeathStarBench Social Network application with three root operations under varying load and workload mixes. Across scenarios, ADASCALE consistently meets SLO targets and improves both latency and throughput. Compared with NetMARKS_Scale, it achieves up to 1.56x, 1.93x, and 1.34x lower average response time for compose-post, read-home-timeline, and read-user-timeline, respectively. It also achieves up to 2.16x, 1.32x, and 1.36x higher throughput for the same operations. These results demonstrate that ADASCALE effectively handles non-stationary traffic, shifts in root operation mix, and heterogeneous communication modes, outperforming a state-of-the-art network-aware autoscaler.

Microservice applications are increasingly deployed across cloud-edge environments, where heterogeneous nodes and time-varying inter-node delays amplify the impact of placement decisions. At the same time, these applications face non-stationary traffic, shifts in the mix of root request operations that exercise different call graphs, and heterogeneous communication modes that determine how network latency and queuing propagate to end-to-end (E2E) performance. Existing autoscalers and network-aware schedulers typically handle only a subset of these dynamics, leading to either compute bottlenecks or inflated cross-node latency and thus SLO violations. The authors propose ADASCALE, an adaptive framework that jointly scales and places microservice replicas under such multi-dimensional dynamics. The core problem addressed is the joint optimization of replica count and placement to satisfy SLOs while minimizing latency and maximizing throughput in dynamic cloud-edge settings.

ADASCALE implements a Monitor-Analyzer-Planner-Executor (MAPE) loop that extracts per-edge and per-service demand from distributed traces and service-mesh metrics, identifies the most critical root operation under a mixed workload, computes SLO-aware replica targets, and then places replicas to minimize a demand-weighted latency objective given the current inter-node latency matrix. To react quickly to networking perturbations, ADASCALE triggers a reactive placement loop, while a steady-state autoscaling loop handles demand shifts.

The results indicate that jointly addressing scaling and placement under multi-dimensional dynamics yields significant improvements over approaches that handle only a subset of these factors. ADASCALE's ability to identify the most critical root operation under a mixed workload allows it to prioritize resources effectively, reducing SLO violations. The reactive placement loop enables quick adaptation to networking perturbations, while the steady-state autoscaling loop handles gradual demand shifts. The demand-weighted latency objective ensures that placement decisions consider both service demand and inter-node latency, leading to better E2E performance. The evaluation on a cloud-edge Kubernetes cluster with DeathStarBench Social Network demonstrates the practical applicability of ADASCALE. However, the study is limited to a specific application and cluster setup; future work could explore other applications and more diverse edge environments. The taxonomy candidates for this work include Architecture, Cybersecurity, Network, and Cryptography, though the primary focus is on architecture and network aspects of microservice scaling and placement.

CS practitioners and researchers

Opening member content…