add perf comparaison script

Author: jpinsonneau

.gitignore

@@ -7,4 +7,5 @@ ebpf-agent.tar
 *.pcap
 protoc/
 release-assets/
+perf/
 

scripts/README_visualize.md

@@ -0,0 +1,90 @@
+# eBPF Performance Visualization Script
+
+## Overview
+
+`visualize_ebpf_performance.py` creates comprehensive visualizations of eBPF agent performance data from CSV files generated by prow.
+
+## Usage
+
+```bash
+python3 scripts/visualize_ebpf_performance.py <csv_file> <prow_id> [--output <output_file>]
+```
+
+### Arguments
+
+- `csv_file`: Path to CSV file containing performance data such as `https://gcsweb-ci.apps.ci.l2s4.p1.openshiftapps.com/gcs/test-platform-results/pr-logs/pull/netobserv_netobserv-ebpf-agent/824/pull-ci-netobserv-netobserv-ebpf-agent-main-ebpf-node-density-heavy-25nodes/1985348508604960768/artifacts/ebpf-node-density-heavy-25nodes/openshift-qe-orion/artifacts/data-netobserv-perf-node-density-heavy-AWS-25w.csv`
+- `prow_id`: Prow ID of the target run to compare against previous runs such as `1985348508604960768`
+- `--output`, `-o`: (Optional) Output PNG file path (default: `perf/ebpf_performance_visualization.png`)
+
+### Examples
+
+```bash
+# Basic usage
+python3 scripts/visualize_ebpf_performance.py data.csv 1985348508604960768
+
+# Specify custom output file
+python3 scripts/visualize_ebpf_performance.py data.csv 1985348508604960768 --output custom_output.png
+
+# With full paths
+python3 scripts/visualize_ebpf_performance.py /path/to/data.csv 1985348508604960768 -o /path/to/output.png
+```
+
+## Requirements
+
+- Python 3
+- matplotlib
+- numpy
+
+Install dependencies:
+```bash
+pip3 install matplotlib numpy
+```
+
+## Output
+
+The script generates an 8-panel visualization (5 rows × 2 columns) showing:
+
+**Row 1:**
+1. **Flows Processed Over Time** - Scatter plot with previous average line
+2. **Flows Processed Comparison** - Bar chart comparing min/avg/max vs current
+
+**Row 2:**
+3. **CPU Usage Over Time** - CPU usage trends with previous average line
+4. **Memory Usage Over Time** - RSS memory usage trends with previous average line
+
+**Row 3:**
+5. **CPU Efficiency** - Flows per minute per core scatter plot with efficiency trend line
+6. **Memory Efficiency** - Flows per minute per MB scatter plot with efficiency trend line
+
+**Row 4:**
+7. **Efficiency Comparison** - Percentage change bar chart (CPU and Memory efficiency vs previous average), full width spanning both columns
+
+**Row 5:**
+8. **Summary Statistics** - Comprehensive text panel split into 3 columns:
+   - **Column 1**: Flows metrics (Flows Processed, Flows Per Minute)
+   - **Column 2**: Resource usage (CPU Usage, Memory RSS)
+   - **Column 3**: Efficiency metrics and Dropped Flows information
+
+## Metrics Visualized
+
+- **Flows Processed**: Total flows and flows per minute
+- **CPU Usage**: Resource consumption (cores) over time
+- **Memory Usage**: RSS memory consumption (GB) over time
+- **Efficiency Metrics**:
+  - **CPU Efficiency**: Flows per minute per core (scatter plot + percentage change)
+  - **Memory Efficiency**: Flows per minute per MB (scatter plot + percentage change)
+- **Dropped Flows**: Reliability indicator (zero is ideal) - shown in summary panel only
+
+## Notes
+
+- The script compares the target prow ID against all previous runs in the CSV
+- Efficiency calculations use rate-based metrics (flows per minute) for accurate comparison
+- Memory efficiency is calculated per MB (not GB) for more granular analysis
+- Efficiency comparison shows percentage change from previous average (green = improvement, orange = regression)
+- All plots include trend lines and reference averages for easy comparison
+- Dropped flows information is displayed in the summary panel (not as a separate plot)
+- Summary statistics are organized into 3 columns for better readability
+- Output is saved at 300 DPI for high-quality visualization
+- By default, output is saved to `perf/` folder (created automatically if it doesn't exist)
+- The `perf/` folder is ignored by git (added to `.gitignore`)
+