wangyuyue/simpyssd

How wangyuyue/simpyssd works

SimpySSD models BeaconGNN workloads executing with in-storage graph sampling on configurable SSD, system, and accelerator designs. The principal workflow starts in the experiment runner: it selects a workload and configurations, builds a graph and sampled subgraphs, runs GNN requests through an event-driven SSD/system model, then collects and plots latency, utilization, speedup, and energy results. The graph preserves documented experiment and analysis stages; implementation links are limited to relationships visible in excerpts or the SOURCE INDEX.

Main entry points: main.py Experiment Runner · system.py System Model · accel_config.py Accelerator Configuration · gnn.py GNN Execution

Components 25

Workload Modeling

Experiment Runner src/main.py
Graph Workloads src/graph.py
Subgraph Sampling src/subgraph.py
GNN Execution src/gnn.py
Accelerator Topology src/gnn_topology.py

Device Simulation

Event Engine src/sim.py
SSD Model src/ssd_estimate.py
System Model src/system.py
DRAM Model src/dram.py
PCIe Model src/pcie.py
Accelerator Model src/nn_accel.py

Design Configuration

SSD Configuration src/ssd_config.py
System Configuration src/system_config.py
Accelerator Configuration src/accel_config.py
SSD Profiles configs/ssd
System Profiles configs/system
Accelerator Profiles configs/accelerator

Results Analysis

Simulation Statistics src/ssd_statistics/sys_stat.py
Energy Estimation src/ssd_statistics/energy_estimate.py
Latency Analysis src/ssd_statistics/dataframe.py
Result Export src/ssd_statistics/dump.py
Result Plotting src/ssd_statistics/plot.py
Series Processing src/ssd_statistics/series_process.py

Other components

Experiment User
Scale-Sim Tool

Connections 30

Experiment User → Experiment Runner: starts experiments README.md
Experiment Runner → Graph Workloads: builds workload src/main.py
Experiment Runner → Subgraph Sampling: samples subgraphs src/main.py
Experiment Runner → GNN Execution: runs workload src/main.py
GNN Execution → Graph Workloads: queries graph src/gnn.py
GNN Execution → SSD Model: issues commands src/gnn.py
Experiment Runner → System Model: creates system src/main.py
System Model → SSD Model: instantiates SSD src/system.py
System Model → PCIe Model: uses PCIe src/system.py
System Model → Accelerator Model: uses accelerator src/system.py
System Model → DRAM Model: uses DRAM src/system.py
SSD Model → Event Engine: schedules events src/ssd_estimate.py
Experiment Runner → Event Engine: executes events src/main.py
Experiment Runner → SSD Configuration: loads profile src/main.py
Experiment Runner → System Configuration: selects profile src/main.py
Experiment Runner → Accelerator Configuration: sets accelerator src/main.py
SSD Configuration → SSD Profiles: reads settings src/ssd_config.py
System Configuration → System Profiles: reads settings src/system_config.py
Accelerator Configuration → Accelerator Profiles: reads settings
Accelerator Configuration → Accelerator Topology: builds topology src/accel_config.py
Accelerator Configuration → Scale-Sim Tool: simulates compute src/accel_config.py
Experiment Runner → Simulation Statistics: records statistics src/main.py
Experiment Runner → Energy Estimation: computes energy src/main.py
Experiment Runner → Result Export: exports results src/main.py
Experiment Runner → Result Plotting: plots results src/main.py
Result Export → Latency Analysis: builds dataframes src/ssd_statistics/dump.py
Result Plotting → Series Processing: processes series src/ssd_statistics/plot.py
Result Plotting → Latency Analysis: plots dataframes src/ssd_statistics/plot.py
Energy Estimation → SSD Configuration: uses SSD settings src/ssd_statistics/energy_estimate.py
Energy Estimation → System Configuration: uses system settings src/ssd_statistics/energy_estimate.py

More Python diagrams: breakthrough/pyscenedetect · ashleve/lightning-hydra-template · defectdojo/django-defectdojo · apple/coremltools · llm-d/llm-d

Drawn by GitDiagram on September 30, 2026 from the repository's file tree, README and source files. Read as Markdown ·