fyp-onboarding

A project to familiarise benchmarking on vhive, knative.

Directory Guide

ª Dockerfile (For Dockerising Worker only) ª go.mod ª go.sum ª hellotest.go (Simple Connection Test Script) ª README.md ª worker.proto (protobuf) ª
+---knative ª worker-service.yaml (Knative Service Manifest) ª
+---loadgen ª ª load.log ª ª loadgen (Compiled binary of load_generator.go) ª ª load_generator.go (Main Load Generator script) ª ª
ª +---logs +---loadgen_basic ª load_gen_basic.go (Simple load generator - sends one request. For debug purposes) ª
+---worker ª worker.go (Main Worker script) ª
+---workerpb (client/server interface) worker.pb.go worker_grpc.pb.go

Deployment Guide

1. Setup a single cluster node according to the vHive Quickstart Guide or use the provided setup script.
2. Clone this repository into the node.
3. If the worker image is not pushed into Docker Hub (or another registry), follow steps 4–6. (Ensure Docker is installed: sudo apt install docker.io) Otherwise, skip to step 7.
4. Build the image: docker build -t <userid>/worker:latest -f Dockerfile .
5. Log in to Docker: docker login -u <username>
6. Push the image into the registry: docker push <userid>/worker:latest
7. Deploy into Knative: kubectl apply -f knative/worker-service.yaml
8. Check if the worker is ready: kubectl get ksvc worker
9. Get the Knative service URL (endpoint). Default external port is 80.
10. Run the Load Generator (replace <URL:80> with the worker endpoint): go run loadgen/load_generator.go --worker=<URL:80>
11. The Load Generator runs and saves output in the /logs folder. It measures requests and end-to-end latency (E2E).