Verification Tool Configuration

Use a YAML file to configure the test.

Example YAML

name: cluster-validation-suite
global:
duration: 60
parameters:
    image-pull-secret: ngc-secret
    image-name: nvcr.io/nvidia/holoscan-for-media/h4m-verification-tool-internals:0.1.0
monitoring:
    - name: SR-IOV
    module: sriov
    - name: PTP
    module: ptp
    - name: GPU
    module: gpu

tests:
- name: Platform Checks
    description:
    Capture and verify system details needed for optimal performance including
    PCIe topology, NIC firmware settings, and kernel arguments
    module: Platform
    enabled: true
    validation:
    nic_config:
        - REAL_TIME_CLOCK_ENABLE: True(1)
        - VF_VPD_ENABLE: True(1)
    sriov_config:
        - sriov: true
    acs_config:
        - acs: false
    kernel_args:
        - default_hugepagesz=2M
        - hugepagesz=2M
        - nosmt
        - nosoftlockup
        - idle=poll
        - cpufreq.default_governor=performance
    gpu_config:
        - pcie_link_width: 16x
    duration: 0
    monitoring: null
    parameters: null

- name: Rivermax Generic API
    description:
    Send data with Rivermax Generic API using kernel bypass to capture maximum
    throughput possible on the cluster
    module: generic_sendx_receive_test
    enabled: true

- name: Rivermax Generic API with GPU
    description:
    Send data with Rivermax Generic API using GPUDirect to capture maximum
    throughput possible on the cluster
    module: generic_sendx_receive_test
    enabled: true
    parameters:
    gpu: 1

- name: Rivermax Media 1080p60
  description:
  Send ST2110 1080p 60 fps video streams with Rivermax Media API
  module: media_sendx_rtp_receive_test
  enabled: true
  parameters:
  streams: 71

- name: Rivermax Media 1080p60 with GPU
    description:
    Send ST2110 1080p 60 fps video streams with Rivermax Media API using GPU
    Direct
    module: media_sendx_rtp_receive_test
    enabled: true
    parameters:
    streams: 71
    gpu: 1

- name: Rivermax Media 4K
    description: Send ST2110 4K 60 fps video streams using Rivermax Media API
    module: media_sendx_rtp_receive_test
    enabled: true
    parameters:
    streams: 17
    width: 3840
    height: 2160

- name: Rivermax Media 4K with GPU
    description:
    Send ST2110 4K 60 fps video streams with Rivermax Media API using GPU
    Direct
    module: media_sendx_rtp_receive_test
    enabled: true
    parameters:
    streams: 17
    width: 3840
    height: 2160
    gpu: 1

Global Parameters

The global section contains settings that apply to all tests in the suite.

Setting	Description	Type	Possible Values
duration	The duration of the test in seconds.	Integer	Any positive integer, for example, 1, 3, or 5.
parameters	The parameters for the test. See Parameters section below.	Dict	Example: image-pull-secret: "" cpu: 4
monitoring	The monitoring for the test. See Monitoring section below.	List	Example: - name: SR-IOV module: sriov

Parameters

Setting	Description	Type	Default Value	Possible Values
image-pull-secret	The image pull secret for the test	String	ngc-secret	The name of the image pull secret to use for the test
image-name	The name of the image to use for the test	String	nvcr.io/nvidia/holoscan-for-media/h4m-verification-tool-internals:0.1.0	The name of the image to use for the test
hugepages-type	The type of hugepages to use for the test	String	hugepages-2Mi	hugepages-2Mi hugepages-1Gi
hugepages	The amount of hugepages to use for the test	String	4Gi	The amount of hugepages to use for the test
memory	The amount of memory to use for the test	String	4Gi	The amount of memory to use for the test
cpu	The number of CPU cores to use for the test	Integer	4	Any positive integer, for example, 1, 2, 4.
gpu	The number of GPUs to use for the test	Integer	None	Any positive integer, for example, 1, 2, 4. If not specified, the test will use no GPUs.
network	The name of the network to use for the test	String	media-a-tx-net
scheduler-name	The name of the scheduler to use for the test	String	topo-aware-scheduler
rivermax-license-secret	The name of the Rivermax license secret to use for the test	String	rivermax-license

Monitoring

Setting	Description	Type	Possible Values
name	The name of the monitoring module.	String	Monitoring module name
module	The module of the monitoring module.	String	sriov ptp gpu

Tests

Platform Checks

This test validates the system configuration including NIC settings, SR-IOV networking configuration, and kernel parameters.

Setting	Description	Type	Possible Values
name	The name of the test.	String
module	Name of the module.	String	Platform
enabled	Whether the test is enabled.	Boolean	true or false
description	Description of the test.	String
validation	Validation for the test.	Dict	nic_config sriov_config acs_config kernel_args gpu_config
duration	The duration of the test.	Integer	0
monitoring	Monitoring for the test.	List	null
parameters	Parameters for the test.	Dict	null

nic_config

nic_config:
  - REAL_TIME_CLOCK_ENABLE: True(1)
  - VF_VPD_ENABLE: True(1)

sriov_config

sriov_config:
  - sriov: true

acs_config

acs_config:
  - acs: false

kernel_args

kernel_args:
  - default_hugepagesz=2M
  - hugepagesz=2M
  - nosmt
  - nosoftlockup
  - idle=poll
  - cpufreq.default_governor=performance

gpu_config

gpu_config:
  - pcie_link_width: 16x

Rivermax Xstream Generic API

Tests basic Rivermax Xstream functionality for sending and receiving generic streams.

- name: Rivermax Generic API
  description:
    Send data with Rivermax Generic API using kernel bypass to capture maximum
    throughput possible on the cluster
  module: generic_sendx_receive_test
  enabled: true
  parameters:
    sender:
      - flows: 1
    receiver:
      - data-size: 1500

parameters

Setting	Description	Type	Default Value
dst-ip	Destination IP address.	String	234.5.6.7
port	Destination port.	Integer	50000
sender	Sender parameters.	Dict
receiver	Receiver parameters.	Dict

sender

Setting	Description	Type	Default Value
flows	Number of flows.	Integer	1
streams	Number of total streams.	Integer	1
threads	Number of threads.	Integer	1
chunks	Number of memory chunks.	Integer	1
packets	Number of packets in chunk.	Integer	4096
payload-size	The packet’s payload size.	Integer	1460
rate-bps	Rate limit in bits per second per flow.	Integer	5000000
rate-burst	The maximum allowed burst size in number of chunks per flow.	Integer	100
no-of-app-cores	The number of application cores.	Integer	null
node-list	The list of nodes to use for the test.	List[int]	null

receiver

Setting	Description	Type	Default Value
data-size	User data (payload) size	Integer	1500
packets	Number of packets to allocate memory for.	Integer	65536
flow-id	Flow ID to use for the test.	Integer	1
wait-event	Wait for packets instead of busy loop	Boolean	true
no-of-app-cores	The number of application cores.	Integer	null
node-list	The list of nodes to use for the test.	List[int]	null

validation

Setting	Description	Type	Default Value
target-gbps	Target Gbps	Float	Calculated at runtime from NIC speed
tolerance-pct	Tolerance percentage	Integer	10

Rivermax Xstream Media API

This test validates Rivermax functionality with GPUDirect support for sending and receiving generic streams.

parameters

- name: Rivermax Media 4K
  description: Send ST2110 4K 60 fps video streams using Rivermax Media API
  module: media_sendx_rtp_receive_test
  enabled: true
  parameters:
    streams: 17
    width: 3840
    height: 2160
    sender:
      - threads: 1
    receiver:
      - threads: 1

Setting	Description	Type	Default Value
dst-ip	Destination IP address.	String	234.5.6.7
port	Destination port.	Integer	50000
width	Width of the stream.	Integer	1920
height	Height of the stream.	Integer	1080
fps	FPS of the stream.	Integer	60
streams	Number of streams.	Integer	1
sender	Sender parameters.	Dict
receiver	Receiver parameters.	Dict

sender

Setting	Description	Type	Default Value
threads	Number of threads to use for sending.	Integer	3
sleep	Application doesn’t block on send	Boolean	false
no-of-app-cores	The number of application cores.	Integer	null
node-list	The list of nodes to use for the test.	List[int]	null

receiver

Setting	Description	Type	Default Value
threads	Number of threads to use for receiving.	Integer	1
streams	Number of streams to receive.	Integer	1
payload-size	Packet’s payload size	Integer	1200
app-hdr-size	Packet’s application header size	Integer	20
ext-seq-num	Parse extended sequence number from RTP payload	Boolean	true
no-of-app-cores	Number of application cores.	Integer	null
node-list	The list of nodes to use for the test.	List[int]	null

validation

Setting	Description	Type	Default Value
target-gbps	Target Gbps	Float	Calculated at runtime from the number of streams, width, height, and FPS
tolerance-pct	Tolerance percentage	Integer	10
dropped-packets-threshold	Max number of dropped packets per second for a partial passed test	Integer	30