Run your first experiment#

Run a small file I/O experiment locally and find its results. This path needs no GPU or scheduler.

Prepare#

Install CloudAI and run the following commands from the repository root. Install the Fio executable; on Ubuntu, use sudo apt-get install fio.

cp conf/common/system/standalone_system.toml standalone.toml

Run#

The example writes an 80 MiB file at /tmp/cloudai-fio-quickstart. Choose another filename in the scenario if you want to measure a different filesystem.

uv run cloudai run --system-config standalone.toml \
    --test-scenario conf/experimental/quickstart/fio.toml
View the scenario
name = "quickstart-fio"

[[Tests]]
id = "fio"
name = "fio"
description = "Small file I/O benchmark"
test_template_name = "Fio"
num_nodes = 1
time_limit = "00:05:00"

  [Tests.cmd_args]
  fio_binary = "fio"

    [Tests.cmd_args.args]
    name = "cloudai-quickstart"
    filename = "/tmp/cloudai-fio-quickstart"
    rw = "randwrite"
    bs = "128k"
    size = "80m"
    iodepth = 1
    numjobs = 1
    group_reporting = true
    thread = true

Inspect#

Open the results directory printed by CloudAI. Inspect experiment.json for execution status and recorded metrics, then the workload output for bandwidth and IOPS. A successful process alone does not establish that you measured the intended device: check the configured path, I/O pattern, block size, and file size. See Read results and generate reports for result interpretation.

Adapt#

Change one Fio setting at a time and compare results under the same conditions. Use Understand experiment configuration to understand the two configuration files. For GPU or multi-node experiments, choose a workload from Workloads and follow Set up your execution environment. For configuration search, continue to Run a configuration search.

On this page