Data Center Configuration File¶
This file defines one or more data center entries under the top-level datacenters key. Each entry describes the compute, memory, and geographic characteristics of a single data center. Below is an example structure; see the field descriptions that follow.
datacenters:
- dc_id: 1
location: "US-CAL-CISO"
timezone_shift: -7
population_weight: 0.18
total_cores: 50000
total_gpus: 1000
total_mem: 80000
dc_config_file: "configs/dcs/dc_config.json"
- dc_id: 2
location: "DE-LU"
timezone_shift: 1
population_weight: 0.22
total_cores: 85000
total_gpus: 600
total_mem: 80000
dc_config_file: "configs/dcs/dc_config.json"
- dc_id: 3
location: "CL-SIC"
timezone_shift: -5
population_weight: 0.20
total_cores: 110000
total_gpus: 300
total_mem: 60000
dc_config_file: "configs/dcs/dc_config.json"
- dc_id: 4
location: "SG"
timezone_shift: 8
population_weight: 0.25
total_cores: 15000
total_gpus: 700
total_mem: 50000
dc_config_file: "configs/dcs/dc_config.json"
- dc_id: 5
location: "AU-NSW"
timezone_shift: 11
population_weight: 0.15
total_cores: 25000
total_gpus: 300
total_mem: 60000
dc_config_file: "configs/dcs/dc_config.json"
Field Reference¶
datacenters (list) : A sequence of data center definitions.
For each item in the list:
dc_id (integer) Unique identifier for the data center.
location (string) A short code representing the geographic region (e.g., “US-CAL-CISO”, “DE-LU”, “SG”).
timezone_shift (integer) Offset in hours from UTC for local time alignment (e.g., -7 for PDT, +1 for CET).
population_weight (float) Relative weight reflecting the fraction of total user population served by this data center.
total_cores (integer) Total number of CPU cores available in the data center.
total_gpus (integer) Total number of GPU devices available.
total_mem (integer) Total memory capacity in GB.
dc_config_file (string) Path to an additional JSON file containing advanced or region-specific parameters (e.g., rack layouts, power limits).
Usage¶
Place your datacenters.yaml in the configs/ directory.
Read this file via the load_yaml loader in train_rl_agent.py.
See also Main Configuration File for how to integrate this configuration into the main environment setup.