NIXL CTPerf#

This workload (test_template_name is NixlPerftest) runs NIXL performance testing suite for comprehensive network performance evaluation.

Usage Examples#

Test TOML example:

name = "my_nixl_perftest_test"
description = "Example NIXL Perftest test"
test_template_name = "NixlPerftest"

[cmd_args]
docker_image_url = "<docker container url here>"
subtest = "sequential-ct-perftest"
num_user_requests = 1
batch_size = 1
num_prefill_nodes = 1
num_decode_nodes = 1
prefill_tp = 4
decode_tp = 4
isl_mean = 10000
isl_scale = 3000
model = "deepseek-r1-distill-llama-70b"

Test Scenario example:

name = "nixl-perftest-test"

[[Tests]]
id = "perftest.1"
num_nodes = 2
time_limit = "00:20:00"

test_name = "my_nixl_perftest_test"

Test-in-Scenario example:

name = "nixl-perftest-test"

[[Tests]]
id = "perftest.1"
num_nodes = 2
time_limit = "00:20:00"

name = "my_nixl_perftest_test"
description = "Example NIXL Perftest test"
test_template_name = "NixlPerftest"

  [Tests.cmd_args]
  docker_image_url = "<docker container url here>"
  subtest = "sequential-ct-perftest"
  num_user_requests = 100
  batch_size = 1
  num_prefill_nodes = 1
  num_decode_nodes = 1
  prefill_tp = 8
  decode_tp = 8
  model = "deepseek-r1-distill-llama-70b"

  [Tests.extra_env_vars]
  CUDA_VISIBLE_DEVICES = "$SLURM_LOCALID"

API Documentation#

Command Arguments#

pydantic model cloudai.workloads.nixl_perftest.nixl_perftest.NixlPerftestCmdArgs[source]#

CmdArgs for NIXL Perftest.

field subtest: Literal['sequential-ct-perftest'] [Required]#

field perftest_script: str = '/workspace/nixl/benchmark/kvbench/main.py'#

field matgen_script: str = '/workspace/nixl/benchmark/kvbench/test/inference_workload_matgen.py'#

field python_executable: str = 'python'#

field num_user_requests: int | list[int] [Required]#

field batch_size: int | list[int] [Required]#

field num_prefill_nodes: int | list[int] [Required]#

field num_decode_nodes: int | list[int] [Required]#

field isl_mean: int | list[int] | None = None#

field isl_scale: int | list[int] | None = None#

field prefill_tp: int | list[int] = 1#

field prefill_pp: int | list[int] = 1#

field prefill_cp: int | list[int] = 1#

field decode_tp: int | list[int] = 1#

field decode_pp: int | list[int] = 1#

field decode_cp: int | list[int] = 1#

field model: str | list[str] | None = None#

field hidden_size: int | None = None#

field num_layers: int | None = None#

field num_heads: int | None = None#

field num_kv_heads: int | None = None#

field dtype_size: int | None = None#

field matgen_args: MatgenCmdArgs [Optional]#

field docker_image_url: str [Required]#: URL of the Docker image to use for the benchmark.

field etcd_path: str = 'etcd'#: Path to the etcd executable.

field wait_etcd_for: int = 60#: Number of seconds to wait for etcd to become healthy.

field etcd_image_url: str | None = None#: Optional URL of the Docker image to use for etcd, by default etcd will be run from the same image as the benchmark.

Test Definition#

class cloudai.workloads.nixl_perftest.nixl_perftest.NixlPerftestTestDefinition( *, name: str, description: str, test_template_name: str, cmd_args: NixlPerftestCmdArgs, extra_env_vars: dict[str, str | List[str]] = {}, extra_cmd_args: dict[str, str] = {}, extra_container_mounts: list[str] = [], git_repos: list[GitRepo] = [], nsys: NsysConfiguration | None = None, predictor: PredictorConfig | None = None, agent: str = 'grid_search', agent_steps: int = 1, agent_metrics: list[str] = ['default'], agent_reward_function: str = 'inverse', agent_config: dict[str, Any] | None = None, )[source]#

Bases: NIXLBaseTestDefinition[NixlPerftestCmdArgs]

TestDefinition for NixlPerftest.