Add Streamlined LLM Serving and Benchmarking Setup Scripts

This PR introduces two new setup scripts that simplify the process of deploying LLM serving infrastructure and benchmarking environments with real-time metrics monitoring.

New Scripts

1. setupGpuVM.sh

A comprehensive setup script for GPU VMs that automates the entire LLM serving stack deployment in a single command.

Features:

• Installs vLLM framework
• Deploys and serves specified LLM model
• Configures GPU telemetry for metrics collection

2. setupBenchmarkVM.sh

Sets up a dedicated benchmark VM with monitoring infrastructure and testing tools.

Features:

• Deploys Prometheus + Grafana monitoring stack for multi-GPU VM monitoring
• Downloads export_metrics.sh for CSV-based metrics export
• Downloads run_guidellm.sh for benchmark execution

Usage Examples

GPU VM Remote Command Execution

Step 1: Install GPU driver (requires reboot - separate step)

# Use existing driver installation remote command
curl -fsSL https://raw.githubusercontent.com/cloud-barista/cb-tumblebug/main/scripts/usecases/llm/installGpuDriver.sh | bash

Step 2: Deploy vLLM, serve model, and configure telemetry

curl -fsSL https://raw.githubusercontent.com/cloud-barista/cb-tumblebug/main/scripts/usecases/llm/setupGpuVMForLLM.sh | bash -s -- Qwen/Qwen2.5-1.5B-Instruct

Benchmark VM Remote Command Execution

Step 1: Setup monitoring stack and download benchmark tools

GPU_VM_IPS='<GPU_VM_IP1> [GPU_VM_IP2] ...'
curl -fsSL https://raw.githubusercontent.com/cloud-barista/cb-tumblebug/main/scripts/usecases/llm/telemetry/setupBenchmarkVM.sh | bash -s -- $GPU_VM_IPS