How msvcbench/mubench works
Project Type and Purpose
• µBench is an orchestration‐and‐benchmarking tool for cloud/edge microservice applications. It is not a monolithic app, but a factory that programmatically generates, deploys, and monitors dummy microservice workloads on Kubernetes.
File‐Structure Analysis
• Top‐level “Deployers”, “ServiceGraphGenerator”, “WorkModelGenerator”, “Benchmarks”, “Autopilots”: these are Python CLI modules that drive the workflow.
• “Configs”: JSON parameter files that configure graph topology, resource settings, traffic patterns, etc.
• “Add-on” and “Monitoring”: optional, complementary scripts and manifests for Horizontal Pod Autoscaler, Istio service mesh, Prometheus/Grafana/Kiali/Jaeger.
• “ServiceCell”: a Dockerized microservice template implementing both REST and gRPC, plus stress‐functions library.
• CI/CD (“.github”), Dockerfile, and Docker-README for containerizing the entire µBench tool.
Key README Insights
• Users run scripts inside a Docker container to generate service graphs & work models, then deploy to Kubernetes via the “K8sDeployer”.
• The deployed application consists of an NGINX gateway (“gw-nginx”) fronting N service‐cell pods, wired by the generated service graph.
• A full observability stack (Prometheus, Grafana, Istio, Kiali, Jaeger) can be installed and configured.