smallfile workload plugin for Arcaflow

arca-smallfile is a workload plugin of the smallfile benchmark tool using the Arcaflow python SDK.

Supported smallfile parameters are defined in the SmallfileParams schema of the smallfile_schema.py file. You define your test parameters in a YAML file to be passed to the plugin command as shown in smallfile-example.yaml.

Image Building

You can change this plugin's image version tag in .github/workflows/carpenter.yaml by editing the IMAGE_TAG variable, and pushing that change to the branch designated in that workflow.

To run directly without the Arcaflow engine:

In order to run the arca-smallfile plugin follow these steps:

Containerized

1. Clone this repository
2. Create the container with docker build -t arca-smallfile <clone_directory>
3. Run cat smallfile-example.yaml | docker run -i arca-smallfile -f -

Native

Prerequisite: smallfile should already be installed on your system.

1. Clone this repository
2. Create a venv in the current directory with python3 -m venv ./venv
3. Activate the venv by running source ./venv/bin/activate
4. Run pip install -r requirements.txt
5. Edit smallfile_plugin.py to set smallfile_dir=<path_to_smallfile_dir>
6. Run ./smallfile_plugin.py -f smallfile-example.yaml

smallfile workload (workload)

Run the smallfile workload with the given parameters

Input

Type:	scope
Root object:	WorkloadParams
Properties	cleanup (bool) Name: Cleanup files Description: Whether to cleanup files and directories after a run Required: No Default (JSON encoded): true Type: bool smallfile_params (reference[SmallfileParams]) Required: Yes Type: reference[SmallfileParams] Referenced object: SmallfileParams
Objects SmallfileParams (object) Type: object Properties auto-pause (enum[string]) Name: Auto pause Description: Auto-adjust the pause time between files Required: No Type: enum[string] Values N: NO Y: YES cleanup-delay-usec-per-file (int) Name: Cleanup delay (usec) Description: Delay in usec after cleanup operation Required: No Type: int dirs-per-dir (int) Name: Directories per directory Description: Maximum number of subdirectories per directory Required: No Type: int file-size (int) Name: File size (B) Description: Size in bytes of files Required: No Type: int files (int) Name: File count Description: Number of files per thread Required: No Type: int files-per-dir (int) Name: Files per directory Description: Maximum number of files per directory Required: No Type: int finish (enum[string]) Name: Finish operations Description: Complete all file operations even if measurement has finished Required: No Type: enum[string] Values N: NO Y: YES fsync (enum[string]) Name: fsync Description: Insert fsync() call before closing a file Required: No Type: enum[string] Values N: NO Y: YES hash-into-dirs (enum[string]) Name: Hash into directories Description: Assign next file to a directory using a hash function Required: No Type: enum[string] Values N: NO Y: YES incompressible (enum[string]) Name: Incompressible files Description: Generate pure-random files that are not compressible Required: No Type: enum[string] Values N: NO Y: YES operation (enum[string]) Name: Operation Description: Type of smallfile operation Required: Yes Type: enum[string] Values append: APPEND chmod: CHMOD cleanup: CLEANUP create: CREATE delete: DELETE delete_renamed: DELETE_RENAMED getxattr: GETXATTR ls-l: LS_L mkdir: MKDIR overwrite: OVERWRITE read: READ readdir: READDIR rename: RENAME rmdir: RMDIR setxattr: SETXATTR stat: STAT swift-get: SWIFT_GET swift-put: SWIFT_PUT symlink: SYMLINK truncate-overwrite: TRUNCATE_OVERWRITE prefix (string) Name: File prefix Description: Filename prefix Required: No Type: string record-size (int) Name: Record size (KB) Description: Size in KB of data transferred in a single system call Required: No Type: int same-dir (enum[string]) Name: Same directory Description: Threads share a single directory Required: No Type: enum[string] Values N: NO Y: YES stonewall (enum[string]) Name: Stonewall Description: Measure thread throughput when another thread has finished Required: No Type: enum[string] Values N: NO Y: YES suffix (string) Name: File suffix Description: Filename suffix Required: No Type: string threads (int) Name: Threads Description: Number of workload threads Required: No Type: int top (string) Name: Top directory Description: Directory where file operations are performed Required: Yes Type: string verify-read (enum[string]) Name: Verify read Description: Verify read data is correct Required: No Type: enum[string] Values N: NO Y: YES xattr-count (int) Name: xattr count Description: Number of extended attributes per file Required: No Type: int xattr-size (int) Name: xattr size (B) Description: Size in bytes of extended attribute value in bytes Required: No Type: int WorkloadParams (object) Type: object Properties cleanup (bool) Name: Cleanup files Description: Whether to cleanup files and directories after a run Required: No Default (JSON encoded): true Type: bool smallfile_params (reference[SmallfileParams]) Required: Yes Type: reference[SmallfileParams] Referenced object: SmallfileParams

Outputs

error

Type:	scope
Root object:	WorkloadError
Properties	error (string) Required: Yes Type: string
Objects WorkloadError (object) Type: object Properties error (string) Required: Yes Type: string

success

Type:	scope
Root object:	WorkloadResults
Properties	sf_params (reference[SmallfileOutputParams]) Required: Yes Type: reference[SmallfileOutputParams] Referenced object: SmallfileOutputParams sf_results (reference[SmallfileOutputResults]) Required: Yes Type: reference[SmallfileOutputResults] Referenced object: SmallfileOutputResults sf_rsptimes (list[reference[SmallfileOutputRsptimes]]) Required: Yes Type: list[reference[SmallfileOutputRsptimes]] List items Type: reference[SmallfileOutputRsptimes] Referenced object: SmallfileOutputRsptimes
Objects SmallfileOutputParams (object) Type: object Properties auto_pause (bool) Name: Auto pause Description: Auto-adjust the pause time between files Required: Yes Type: bool cleanup_delay_usec_per_file (int) Name: Cleanup delay (usec) Description: Delay in usec after cleanup operation Required: Yes Type: int file_size (int) Name: File size (B) Description: Size in bytes of files Required: Yes Type: int file_size_distr (int) Name: File size distribution Description: Always '-1' as unsupported feature Required: Yes Type: int files_per_dir (int) Name: Files per directory Description: Maximum number of files per directory Required: Yes Type: int files_per_thread (int) Name: File count Description: Number of files per thread Required: Yes Type: int finish_all_requests (enum[string]) Name: Finish operations Description: Complete all file operations even if measurement has finished Required: Yes Type: enum[string] Values N: NO Y: YES fname_prefix (string) Name: File prefix Description: Filename prefix Required: No Type: string fname_suffix (string) Name: File suffix Description: Filename suffix Required: No Type: string fsync_after_modify (enum[string]) Name: fsync Description: Insert fsync() call before closing a file Required: Yes Type: enum[string] Values N: NO Y: YES hash_to_dir (enum[string]) Name: Hash into directories Description: Assign next file to a directory using a hash function Required: Yes Type: enum[string] Values N: NO Y: YES host_set (string) Name: Host set Description: Always 'localhost' because we don't use smallfile's multi-host features Required: Yes Type: string host_timeout (int) Name: Host timeout Description: Always '3' as unsupported feature Required: Yes Type: int launch_by_daemon (bool) Name: Launch by daemon Description: Always 'false' because we don't use smallfile's daemon feature Required: Yes Type: bool min_directories_per_sec (int) Name: Minimum directories per second Description: Always '50' as unsupported feature Required: Yes Type: int network_sync_dir (string) Name: Network sync directory Description: Subdirectory of top for network sync Required: Yes Type: string operation (enum[string]) Name: Operation Description: Type of smallfile operation Required: Yes Type: enum[string] Values append: APPEND chmod: CHMOD cleanup: CLEANUP create: CREATE delete: DELETE delete_renamed: DELETE_RENAMED getxattr: GETXATTR ls-l: LS_L mkdir: MKDIR overwrite: OVERWRITE read: READ readdir: READDIR rename: RENAME rmdir: RMDIR setxattr: SETXATTR stat: STAT swift-get: SWIFT_GET swift-put: SWIFT_PUT symlink: SYMLINK truncate-overwrite: TRUNCATE_OVERWRITE pause_between_files (float) Name: Pause between files Description: Always '0.0' as unsupported feature Required: Yes Type: float permute_host_dirs (enum[string]) Name: Permute host directories Description: Always 'N' as unsupported feature Required: Yes Type: enum[string] Values N: NO Y: YES share_dir (enum[string]) Name: Same directory Description: Threads share a single directory Required: Yes Type: enum[string] Values N: NO Y: YES startup_timeout (int) Name: Startup timeout Description: Always '3' as unsupported feature Required: Yes Type: int stonewall (enum[string]) Name: Stonewall Description: Measure thread throughput when another thread has finished Required: Yes Type: enum[string] Values N: NO Y: YES threads (int) Name: Threads Description: Number of workload threads Required: Yes Type: int top (string) Name: Top directory Description: Directory where file operations are performed Required: Yes Type: string total_hosts (int) Name: Total hosts Description: Total number of hosts in the test Required: Yes Type: int verify_read (enum[string]) Name: Verify read Description: Verify read data is correct Required: Yes Type: enum[string] Values N: NO Y: YES version (string) Name: Version Description: Smallfile version Required: Yes Type: string xattr_count (int) Name: xattr count Description: Number of extended attributes per file Required: Yes Type: int xattr_size (int) Name: xattr size (B) Description: Size in bytes of extended attribute value in bytes Required: Yes Type: int SmallfileOutputResults (object) Type: object Properties IOPS (float) Name: Total IOPS Description: IOPS rate for the job Required: Yes Type: float MiBps (float) Name: Total MiBps Description: MiBps rate for the job Required: Yes Type: float date (string) Name: Date Description: Test run date Required: Yes Type: string elapsed (float) Name: Elapsed time Description: Elapsed time of the job Required: Yes Type: float files (int) Name: Total files Description: Total files for the job Required: Yes Type: int filesPerSec (float) Name: Total files per second Description: Files per second rate for the job Required: Yes Type: float pctFilesDone (float) Name: Percent files done Description: Total percent of file transactions completed Required: Yes Type: float records (int) Name: Total records Description: Total records for the job Required: Yes Type: int startTime (float) Name: Start time Description: Test start time Required: Yes Type: float status (string) Name: Status Description: Test run status Required: Yes Type: string thread (map[int,reference[SmallfileOutputThread]]) Required: Yes Type: map[int,reference[SmallfileOutputThread]] Key type Type: int Value type Type: reference[SmallfileOutputThread] Referenced object: SmallfileOutputThread totalDataGB (float) Name: Total Data (GB) Description: Total data in GB Required: Yes Type: float totalhreads (int) Name: Total threads Description: Total number of threads Required: Yes Type: int SmallfileOutputRsptimes (object) Type: object Properties host_thread (string) Name: Host thread Description: Job ID and thread identifier Required: Yes Type: string max (float) Name: Maximum Description: Maximum response time Required: Yes Type: float mean (float) Name: Mean Description: Mean response time Required: Yes Type: float min (float) Name: Minimum Description: Minimum response time Required: Yes Type: float pctdev (float) Name: Percent deviation Description: Standard deviation as a percent of mean Required: Yes Type: float pctile50 (float) Name: 50th percentile Description: 50th percentile response time Required: Yes Type: float pctile90 (float) Name: 90th percentile Description: 90th percentile response time Required: Yes Type: float pctile95 (float) Name: 95th percentile Description: 95th percentile response time Required: Yes Type: float pctile99 (float) Name: 99th percentile Description: 99th percentile response time Required: Yes Type: float samples (int) Name: Samples Description: Number of file samples Required: Yes Type: int SmallfileOutputThread (object) Type: object Properties IOPS (float) Name: Thread IOPS Description: IOPS rate for the thread Required: Yes Type: float MiBps (float) Name: Thread MiBps Description: MiBps rate for the thread Required: Yes Type: float elapsed (float) Name: Thread elapsed time Description: Elapsed time of the thread Required: Yes Type: float files (int) Name: Thread files Description: Total files for the thread Required: Yes Type: int filesPerSec (float) Name: Thread files per second Description: Files per second rate for the thread Required: Yes Type: float records (int) Name: Thread records Description: Total records for the thread Required: Yes Type: int WorkloadResults (object) Type: object Properties sf_params (reference[SmallfileOutputParams]) Required: Yes Type: reference[SmallfileOutputParams] Referenced object: SmallfileOutputParams sf_results (reference[SmallfileOutputResults]) Required: Yes Type: reference[SmallfileOutputResults] Referenced object: SmallfileOutputResults sf_rsptimes (list[reference[SmallfileOutputRsptimes]]) Required: Yes Type: list[reference[SmallfileOutputRsptimes]] List items Type: reference[SmallfileOutputRsptimes] Referenced object: SmallfileOutputRsptimes