Skip to main content

Flexible I/O Tester (FIO) Workload Profiles

The following profiles run customer-representative or benchmarking scenarios using the Flexible I/O Tester (FIO) workload.

PERF-IO-FIO.json​

Runs an IO-intensive workload using the Flexible IO Tester (FIO) toolset to test performance of disks on the system. Although this profile can run on Windows, it is predominantly used on Linux systems. The profile runs the workload in-parallel on ALL disks that match the "DiskFilter" parameter (see below) by default.

Additonally this profile is designed to auto-scale to the number of cores on the system on which it runs. It uses a simple algorithm to determine 2 key aspects of the workload execution.

  • Total number of jobs/threads = <# of logical cores> / 2

    Examples:
    For a 16-core system:
    16/2 = 16 concurrent jobs/threads per DiskSpd execution (i.e. # threads to run I/O operations against the test file concurrently).

    For a 64-core system:
    64/2 = 32 concurrent jobs/threads per DiskSpd execution (i.e. # threads to run I/O operations against the test file concurrently).

  • Total I/O depth = 512 / < Total number of jobs/threads>

    Examples:
    For a 16 core system:
    total # jobs/threads = 16/2 (above) -> 512/16 = 32

    For a 64 core system:
    total # jobs/threads = 64/2 (above) -> 512/32 = 16

  • Supported Platform/Architectures

    • linux-x64
    • linux-arm64
    • win-x64

  • Supported Operating Systems

    • Ubuntu 18
    • Ubuntu 20
    • Ubuntu 22
    • Ubuntu 24
    • Azlinux 3

  • Supports Disconnected Scenarios

    • No. Internet connection required.

  • Dependencies
    The dependencies defined in the 'Dependencies' section of the profile itself are required in order to run the workload operations effectively.

    • Internet connection.
    • Disk mount points exist for the disks to be targeted. Virtual Client will generally ensure that mount points exist by default. Details for mount point creation procedures can be found in the 'Testing Disks' documentation above.
    • Any 'DiskFilter' parameter value used should match the set of disks desired. See the link for 'Testing Disks' above.

    Additional information on components that exist within the 'Dependencies' section of the profile can be found in the following locations:

  • Scenarios
    The following scenarios are covered by this workload profile.

    • Random Write Operations
      • 4k block size, multiple jobs/threads per disk
      • 8k block size, multiple jobs/threads per disk
      • 12k block size, multiple jobs/threads per disk
      • 16k block size, multiple jobs/threads per disk
      • 1024k block size, multiple jobs/threads per disk
    • Random Read Operations
      • 4k block size, multiple jobs/threads per disk
      • 8k block size, multiple jobs/threads per disk
      • 12k block size, multiple jobs/threads per disk
      • 16k block size, multiple jobs/threads per disk
      • 1024k block size, multiple jobs/threads per disk
    • Sequential Write Operations
      • 4k block size, multiple jobs/threads per disk
      • 8k block size, multiple jobs/threads per disk
      • 12k block size, multiple jobs/threads per disk
      • 16k block size, multiple jobs/threads per disk
      • 1024k block size, multiple jobs/threads per disk
    • Sequential Read Operations
      • 4k block size, multiple jobs/threads per disk
      • 8k block size, multiple jobs/threads per disk
      • 12k block size, multiple jobs/threads per disk
      • 16k block size, multiple jobs/threads per disk
      • 1024k block size, multiple jobs/threads per disk
    • Disk/Data Integrity Verification, Random Write Operations
      • 4k block size, queue depth of 1, single job/thread per disk
      • 16k block size, queue depth of 1, single job/thread per disk
      • 1024k block size, queue depth of 1, single job/thread per disk
    • Disk/Data Integrity Verification, Sequential Write Operations
      • 4k block size, queue depth of 1, single job/thread per disk
      • 16k block size, queue depth of 1, single job/thread per disk
      • 1024k block size, queue depth of 1, single job/thread per disk

  • Profile Parameters
    The following parameters can be optionally supplied on the command line to modify the behaviors of the workload.

    ParameterPurposeDefault Value
    DataIntegrityFileSizeOptional. Defines the size of the file/disk space that will be used for profile disk integrity scenarios/actions.4GB
    DiskFilterOptional. Filter allowing the user to select the disks on which to test.

    See the link 'Testing Disks' at the top for more details.    		BiggestSize
    DiskFillSizeOptional. Allows the user to override the default disk fill size used in the FIO profile (e.g. 500GB -> 26GB). This enables the profile to be used in scenarios where the disk size is very small (e.g. local/temp disk -> 32GB in size).500GB
    DurationOptional. Defines the amount of time to run each FIO scenario/action within the profile.5 minutes
    EngineOptional. Defines the I/O engine to use for the FIO operations (e.g. posixaio, libaio, windowsaio).Linux = libaio, Windows = windowsaio
    FileSizeOptional. Allows the user to override the default file size used in the FIO profile (e.g. 496GB -> 26GB). This enables the profile to be used in scenarios where the disk size is very small (e.g. local/temp disk -> 32GB in size).496GB
    ProcessModelOptional. Defines how the FIO processes will be executed. The following are valid process models:

    SingleProcess
    Executes a single FIO process running 1 job targeting I/O operations against each disk. Results are separated per-disk.

    SingleProcessPerDisk
    Executes a single FIO process for each disk with each process running 1 job targeting I/O operations against that disk (higher stress profile). Results are separated per-disk.

    SingleProcessAggregated
    Executes a single FIO process running 1 job per disk targeting I/O operations against that disk. Results are provided as an aggregation across all disks (i.e. a rollup).    		SingleProcess
    QueueDepthOptional. Defines the I/O queue depth to use for the operations512/the thread count
    ThreadCountOptional. Specifies the number of distinct parallel operations/threads to run per job.# logical processors / 2
    • Profile Component Parameters
      The following section describes the parameters used by the individual components in the profile.
    ParameterPurposeAccepted Values
    ScenarioScenario use to define the given action of profile. This can be used to specify exact actions to run or exclude from the profile.Any string
    MetricsScenarioThe name to use as the "scenario" for all metrics output for the particular profile action.
    CommandLineThe command line parameters for FIO tool set.Any Valid FIO arguments
    DiskFilterFilter allowing the user to select the disks on which to test.See the link 'Testing Disks' at the top for more details.
    DurationDefines the amount of time to run each FIO scenario/action within the profile.integer or time span
    EngineOptional. Defines the I/O engine to use for the FIO operations (e.g. posixaio, libaio, windowsaio).Linux = libaio, Windows = windowsaio
    PackageNameThe logical name for FIO package downloaded and that contains the toolset.
    ProcessModelDefines how the FIO processes will be executed.SingleProcess
    Executes a single FIO process running 1 job targeting I/O operations against each disk. Results are separated per-disk.

    SingleProcessPerDisk
    Executes a single FIO process for each disk with each process running 1 job targeting I/O operations against that disk (higher stress profile). Results are separated per-disk.

    SingleProcessAggregated
    Executes a single FIO process running 1 job per disk targeting I/O operations against that disk. Results are provided as an aggregation across all disks (i.e. a rollup).
    QueueDepthDefines the I/O queue depth to use for the operationsinteger
    TagsTags useful for telemetry dataAny comma-separated string
    ThreadCountSpecifies the number of distinct parallel operations/threads to run per job.

  • Profile Runtimes
    See the 'Metadata' section of the profile for estimated runtimes. These timings represent the length of time required to run a single round of profile actions. These timings can be used to determine minimum required runtimes for the Virtual Client in order to get results. These are often estimates based on the number of system cores.

  • Usage Examples
    The following section provides a few basic examples of how to use the workload profile. See the documentation at the top on 'Testing Disks' for information on how to target select disks on the system.

    # Run the workload on the system (default = largest disks)
    ./VirtualClient --profile=PERF-IO-FIO.json --system=Demo --timeout=1440 --packageStore="{BlobConnectionString|SAS Uri}"

    # The example above runs on the same disks as having DiskFilter=BiggestSize
    ./VirtualClient --profile=PERF-IO-FIO.json --system=Demo --timeout=1440 --packageStore="{BlobConnectionString|SAS Uri}" --parameters=DiskFilter=BiggestSize

    # Run the workload against the operating system disk
    ./VirtualClient --profile=PERF-IO-FIO.json --system=Demo --timeout=1440 --packageStore="{BlobConnectionString|SAS Uri}" --parameters=DiskFilter=OSDisk

    # Run the workload against all of the disks except the operating system disk.
    ./VirtualClient --profile=PERF-IO-FIO.json --system=Demo --timeout=1440 --packageStore="{BlobConnectionString|SAS Uri}" --parameters=DiskFilter=OSDisk:false

    # Run the workload on specific drives/disks
    ./VirtualClient --profile=PERF-IO-FIO.json --system=Demo --timeout=1440 --packageStore="{BlobConnectionString|SAS Uri}" --parameters=DiskFilter=DiskPath:/dev/sdc1,/dev/sdd1

    # Run against smaller disks on the system
    ./VirtualClient --profile=PERF-IO-FIO.json --system=Demo --timeout=1440 --packageStore="{BlobConnectionString|SAS Uri}" --parameters=DiskFilter=OSDisk:false&smallestSize,,,DiskFillSize=26G,,,FileSize=26G

    # Run specific scenarios only. Each action in a profile as a 'Scenario' name.
    ./VirtualClient --profile=PERF-IO-FIO.json --system=Demo --timeout=1440 --packageStore="{BlobConnectionString|SAS Uri}" --scenarios=RandomWrite_4k_BlockSize,RandomWrite_8k_BlockSize,RandomRead_8k_BlockSize,RandomRead_4k_BlockSize

PERF-IO-FIO-OLTP.json​

Runs an IO-intensive workload using the Flexible IO Tester (FIO) toolset. Multi-throughput OLTP-C workload to emulate a SQL Server OLTP disk workload by running four workload compononents in-parallel: random reads, random writes, sequential reads and sequential writes each with an overall weight/percentage.

Examples:
For Total IOPS = 5000, Random Read Weight = 5416, Random Write Weight = 4255, Sequential Read Weight = 0 , Sequential Write Weight = 329
- Random Read IOPS = (5000 * 5416)/(5416+4255+0+329) = 2708
- Random Write IOPS = (5000 * 4255)/(5416+4255+0+329) = 2128
- Sequential Read IOPS = (5000 * 0)/(5416+4255+0+329) = 0
- Sequential Write IOPS = (5000 * 329)/(5416+4255+0+329) = 164

Random IO : It represents the Database of OLTP-C workload. Sequential IO : It represents the logs of OLTP-C workload. Therefore, they are performed on different disks

  • Supported Platform/Architectures

    • linux-x64
    • linux-arm64
    • win-x64

  • Supported Operating Systems

    • Ubuntu 18
    • Ubuntu 20
    • Ubuntu 22
    • Ubuntu 24
    • Azlinux 3

  • Dependencies
    The dependencies defined in the 'Dependencies' section of the profile itself are required in order to run the workload operations effectively.

    • Internet connection.
    • Disk mount points exist for the disks to be targeted. Virtual Client will generally ensure that mount points exist by default. Details for mount point creation procedures can be found in the 'Testing Disks' documentation above.
    • Any 'DiskFilter' parameter value used should match the set of disks desired. See the link for 'Testing Disks' above.

    Additional information on components that exist within the 'Dependencies' section of the profile can be found in the following locations:

  • Profile Parameters
    The following parameters can be optionally supplied on the command line to modify the behaviors of the workload.

    ParameterPurposeDefault Value
    DiskFilterDisk filter to choose disks. Default is to test on biggest non-OS disks.BiggestSize
    RandomIOFileSizeOptional. Allows the user to override the default random io file size used in the profile (e.g. 124GB -> 26GB). This enables the profile to be used in scenarios where the disk size is very small (e.g. local/temp disk -> 32GB in size).124GB
    SequentialIOFileSizeOptional. Allows the user to override the default random io file size used in the profile.20GB
    DirectIOOptional. Set to true to avoid using I/O buffering and to operate directly against the disk. Set to false to use I/O buffering.true
    InitializeDisksInParallelOptional. Specifies whether uninitialized/unformatted disks on the system should be initialized + formatted in parallel.true (initialized in-parallel)
    SequentialDiskCountOptional. Specifies the number of disk that will have Sequential I/O from Selected Disks.1

  • Profile Component Parameters The following section describes the parameters used by the individual components in the profile.

    ParameterPurpose
    DirectIODirect IO parameter for FIO toolset
    DurationSecType of Input Output operation
    JobFilesTemplate job files to be used
    RandomReadBlockSizeRandom read component's Block size. If it is provided it overwrites the DefaultRandomIOBlockSize for Random read component.
    RandomReadNumJobsRandom read component's Number of jobs. If it is provided it overwrites the DefaultNumJobs for Random read component.
    RandomWriteBlockSizeRandom write component's Block size. If it is provided it overwrites the DefaultRandomIOBlockSize for Random write component.
    RandomWriteNumJobsRandom write component's Number of jobs. If it is provided it overwrites the DefaultNumJobs for Random write component.
    SequentialReadBlockSizeSequential read component's Block size. If it is provided it overwrites the DefaultSequentialIOBlockSize for Sequential read component.
    SequentialReadNumJobsSequential read component's Number of jobs. If it is provided it overwrites the DefaultNumJobs for Sequential read component.
    SequentialWriteBlockSizeSequential write component's Block size. If it is provided it overwrites the DefaultSequentialIOBlockSize for Sequential write component.
    SequentialWriteNumJobsSequential write component's Number of jobs. If it is provided it overwrites the DefaultNumJobs for Sequential write component.
    ProcessModelAllows the user to override the default value you can selection Single Process for all disk(SingleProcess) or 1 process for each disk under test (SingleProcessPerDisk).
    ScenarioScenario use to define the given action of profile
    TagsTags usefull for telemetry data

  • Profile Runtimes
    See the 'Metadata' section of the profile for estimated runtimes. These timings represent the length of time required to run a single round of profile actions. These timings can be used to determine minimum required runtimes for the Virtual Client in order to get results. These are often estimates based on the number of system cores.

  • Usage Examples
    The following section provides a few basic examples of how to use the workload profile.

    # Run the workload on the system
    ./VirtualClient --profile=PERF-IO-FIO-OLTP.json --timeout=1440 --packageStore="{BlobConnectionString|SAS Uri}"