harmony/iobench/dash/iobench-dash-v2.py

import dash
from dash import dcc, html, Input, Output
import plotly.express as px
import pandas as pd
import dash_bootstrap_components as dbc
import io

# --- Sample Data ---
# In a real-world scenario, you would load this from a CSV file.
# For this self-contained example, we define the data directly.
# Example: df = pd.read_csv('benchmark_data.csv')
csv_data = """
config,op_type,iops,latency_ms,throughput_mbs
All-HDD,4k_random_read,260,60,1.02
All-HDD,4k_random_write,100,150,0.39
All-HDD,64k_sequential_read,2100,30,131.25
All-HDD,64k_sequential_write,1500,42,93.75
HDD+SSD_WAL,4k_random_read,270,58,1.05
HDD+SSD_WAL,4k_random_write,160,100,0.62
HDD+SSD_WAL,64k_sequential_read,2150,29,134.37
HDD+SSD_WAL,64k_sequential_write,1800,35,112.5
HDD+SSD_WAL_DB,4k_random_read,1250,12,4.88
HDD+SSD_WAL_DB,4k_random_write,1550,10,6.05
HDD+SSD_WAL_DB,64k_sequential_read,2200,28,137.5
HDD+SSD_WAL_DB,64k_sequential_write,2000,32,125
All-NVMe,4k_random_read,400000,0.1,1562.5
All-NVMe,4k_random_write,350000,0.12,1367.18
All-NVMe,64k_sequential_read,16000,4,1000
All-NVMe,64k_sequential_write,12500,5,800
"""

# Read the data using pandas
df = pd.read_csv(io.StringIO(csv_data))

# Initialize the Dash app with a Bootstrap theme
app = dash.Dash(__name__, external_stylesheets=[dbc.themes.FLATLY])

# --- App Layout ---
app.layout = dbc.Container([
    # Header
    dbc.Row([
        dbc.Col([
            html.H1("Ceph Cluster Benchmark Visualizer", className="text-primary"),
            html.P(
                "An interactive tool to compare performance metrics across different Ceph storage configurations.",
                className="lead"
            )
        ])
    ], className="my-4"),

    # Controls and Graphs Row
    dbc.Row([
        # Control Panel Column
        dbc.Col([
            dbc.Card([
                dbc.CardBody([
                    html.H4("Control Panel", className="card-title"),
                    html.Hr(),
                    
                    # Metric Selection Radio Buttons
                    dbc.Label("Select Metric to Display:", html_for="metric-selector"),
                    dcc.RadioItems(
                        id='metric-selector',
                        options=[
                            {'label': 'IOPS (Input/Output Operations Per Second)', 'value': 'iops'},
                            {'label': 'Latency (in Milliseconds)', 'value': 'latency_ms'},
                            {'label': 'Throughput (in MB/s)', 'value': 'throughput_mbs'}
                        ],
                        value='iops', # Default value
                        labelClassName="d-block" # Display labels as blocks
                    ),
                    html.Hr(),

                    # Configuration Selection Checklist
                    dbc.Label("Select Configurations to Compare:", html_for="config-checklist"),
                    dcc.Checklist(
                        id='config-checklist',
                        options=[{'label': config, 'value': config} for config in df['config'].unique()],
                        value=df['config'].unique(), # Select all by default
                        labelClassName="d-block"
                    ),
                ])
            ], className="mb-4")
        ], width=12, lg=4), # Full width on small screens, 1/3 on large

        # Graph Display Column
        dbc.Col([
            dcc.Graph(id='benchmark-graph')
        ], width=12, lg=8) # Full width on small screens, 2/3 on large
    ])
], fluid=True) # Use a fluid container for full-width layout


# --- Callback Function ---
# This function connects the controls to the graph
@app.callback(
    Output('benchmark-graph', 'figure'),
    [Input('metric-selector', 'value'),
     Input('config-checklist', 'value')]
)
def update_graph(selected_metric, selected_configs):
    """
    This function is triggered whenever a control's value changes.
    It filters the dataframe and returns an updated bar chart figure.
    """
    if not selected_configs:
        # If no configs are selected, return an empty figure to avoid errors
        return go.Figure().update_layout(
            title="Please select a configuration to view data.",
            xaxis_title="",
            yaxis_title=""
        )

    # Filter the DataFrame based on the selected configurations
    filtered_df = df[df['config'].isin(selected_configs)]

    # Create the bar chart using Plotly Express
    fig = px.bar(
        filtered_df,
        x='op_type',
        y=selected_metric,
        color='config',
        barmode='group', # Group bars for different configs side-by-side
        labels={
            "op_type": "Benchmark Operation Type",
            "iops": "IOPS (Higher is Better)",
            "latency_ms": "Latency in ms (Lower is Better)",
            "throughput_mbs": "Throughput in MB/s (Higher is Better)",
            "config": "Storage Configuration"
        },
        title=f"Benchmark Comparison for: {selected_metric.replace('_', ' ').title()}",
        height=600 # Set a fixed height for the graph
    )

    # Update layout for better readability
    fig.update_layout(
        xaxis_title="Operation Type",
        yaxis_title=selected_metric.replace('_', ' ').title(),
        legend_title="Configuration",
        title_x=0.5, # Center the title
        xaxis={'categoryorder':'total descending' if selected_metric != 'latency_ms' else 'total ascending'}
    )

    return fig

# --- Run the App ---
if __name__ == '__main__':
    # Use debug=True for development, allowing hot-reloading
    app.run(debug=True)