Convergence of Edge Services & Edge Infrastructure
Anurag Ranjan, Francesc Guim, Mandar M. Chincholkar, Prakash Ramchandran, Ranjan Kumar Mishra, Sunku Ranganath · 2021
An edge cloud is a highly distributed heterogeneous edge infrastructure that follows principles of disaggregated, cloud-native, fully orchestrated, and provisioned architecture, which can manage the lifecycle of application workloads across different vertical domains. Consolidating multiple workloads, such as vision analytics, AI, industrial control, CDN, etc., enables enterprises to onboard various edge services and applications in an optimal manner. To deliver an autonomous architecture capable of deploying edge native instances as a standard workflow, we face challenges around workload onboarding, optimizing services based on underlying hardware capabilities, interoperability, etc. This paper presents key components required in platforms hosting edge services and shows these gains using Smart Edge Open (SE-O), a deployment-ready open platform, as a reference. We selected video analytics and content delivery network (CDN) as workloads to evaluate the platform performance. Based on workload placement with platform awareness and resource optimization, we were able to reduce the performance degradation by 50% in comparison to a manual density-based optimization at the edge.