Workloads

Workload data defines the tasks and power measurements that OpenDT simulates.

Workload Directory

Workloads are stored in the workload/ directory. Each workload has its own subdirectory:

workload/
└── SURF/
    ├── tasks.parquet
    ├── fragments.parquet
    ├── consumption.parquet
    ├── carbon.parquet
    └── topology.json

Required Files

File	Description
tasks.parquet	Task definitions with submission times and resource requirements
fragments.parquet	Task execution profiles showing resource usage over time
consumption.parquet	Actual power measurements from the real datacenter
carbon.parquet	Grid carbon intensity data
topology.json	Datacenter hardware configuration

Selecting a Workload

Set the workload in your configuration file:

services:
  dc-mock:
    workload: "SURF"

The value is the directory name under workload/.

Creating Custom Workloads

To add a new workload:

1. Create a directory under workload/
2. Add the required Parquet files
3. Create a topology.json matching your datacenter hardware
4. Reference the directory name in your configuration

Data Format

tasks.parquet

Column	Type	Description
id	int	Unique task identifier
submission_time	timestamp	When the task was submitted
duration	int	Task duration in milliseconds
cpu_count	int	Number of CPU cores
cpu_capacity	float	CPU speed in MHz
mem_capacity	int	Memory in MB

fragments.parquet

Column	Type	Description
id	int	Fragment identifier
task_id	int	Parent task ID
duration	int	Fragment duration in milliseconds
cpu_count	int	CPUs used
cpu_usage	float	CPU utilization

consumption.parquet

Column	Type	Description
timestamp	timestamp	Measurement time
power_draw	float	Power in Watts
energy_usage	float	Energy in Joules