Red Hat Training

A Red Hat training course is available for Red Hat OpenStack Platform

Chapter 7. Planning your OVS-DPDK deployment

To optimize your Open vSwitch with Data Plane Development Kit (OVS-DPDK) deployment for NFV, you should understand how OVS-DPDK uses the Compute node hardware (CPU, NUMA nodes, memory, NICs) and the considerations for determining the individual OVS-DPDK parameters based on your Compute node.

Important

When using OVS-DPDK and the OVS native firewall (a stateful firewall based on conntrack), you can track only packets that use ICMPv4, ICMPv6, TCP, and UDP protocols. OVS marks all other types of network traffic as invalid.

See NFV performance considerations for a high-level introduction to CPUs and NUMA topology.

7.1. OVS-DPDK with CPU partitioning and NUMA topology

OVS-DPDK partitions the hardware resources for host, guests, and OVS-DPDK itself. The OVS-DPDK Poll Mode Drivers (PMDs) run DPDK active loops, which require dedicated cores. This means a list of CPUs and Huge Pages are dedicated to OVS-DPDK.

A sample partitioning includes 16 cores per NUMA node on dual socket Compute nodes. The traffic requires additional NICs since the NICs cannot be shared between the host and OVS-DPDK.

OpenStack NFV NUMA 9 0219
Note

DPDK PMD threads must be reserved on both NUMA nodes even if a NUMA node does not have an associated DPDK NIC.

OVS-DPDK performance also depends on reserving a block of memory local to the NUMA node. Use NICs associated with the same NUMA node that you use for memory and CPU pinning. Also ensure both interfaces in a bond are from NICs on the same NUMA node.

7.2. Overview of workflows and derived parameters

Important

This feature is available in this release as a Technology Preview, and therefore is not fully supported by Red Hat. It should only be used for testing, and should not be deployed in a production environment. For more information about Technology Preview features, see Scope of Coverage Details.

You can use the OpenStack Workflow (mistral) service to derive parameters based on the capabilities of your available bare-metal nodes. OpenStack workflows use a .yaml file to define a set of tasks and actions to perform. You can use a pre-defined workbook, derive_params.yaml, in the tripleo-common/workbooks/ directory. This workbook provides workflows to derive each supported parameter from the results retrieved from Bare Metal introspection. The derive_params.yaml workflows use the formulas from tripleo-common/workbooks/derive_params_formulas.yaml to calculate the derived parameters.

Note

You can modify the formulas in derive_params_formulas.yaml to suit your environment.

The derive_params.yaml workbook assumes all nodes for a given composable role have the same hardware specifications. The workflow considers the flavor-profile association and nova placement scheduler to match nodes associated with a role and uses the introspection data from the first node that matches the role.

See Troubleshooting Workflows and Executions for details on OpenStack workflows.

You can use the -p or --plan-environment-file option to add a custom plan_environment.yaml file to the openstack overcloud deploy command. The custom plan_environment.yaml file provides the list of workbooks and any input values to pass into the workbook. The triggered workflows merge the derived parameters back into the custom plan_environment.yaml, where they are available for the overcloud deployment. You can use these derived parameter results to prepare your overcloud images.

See Plan Environment Metadata for details on how to use the --plan-environment-file option in your deployment.

7.3. Derived OVS-DPDK parameters

The workflows in derive_params.yaml derive the DPDK parameters associated with the matching role that uses the ComputeNeutronOvsDpdk service.

The following is the list of parameters the workflows can automatically derive for OVS-DPDK:

  • IsolCpusList
  • KernelArgs
  • NovaReservedHostMemory
  • NovaVcpuPinSet
  • OvsDpdkCoreList
  • OvsDpdkSocketMemory
  • OvsPmdCoreList

The OvsDpdkMemoryChannels parameter cannot be derived from the introspection memory bank data since the format of memory slot names are not consistent across different hardware environments.

In most cases, OvsDpdkMemoryChannels should be 4 (default). Use your hardware manual to determine the number of memory channels per socket and use this value to override the default.

See Section 8.1, “Deriving DPDK parameters with workflows” for configuration details.

7.4. Overview of manually calculated OVS-DPDK parameters

This section describes how Open vSwitch with Data Plane Development Kit (OVS-DPDK) uses parameters within the director network_environment.yaml HEAT templates to configure the CPU and memory for optimum performance. Use this information to evaluate the hardware support on your Compute nodes and how best to partition that hardware to optimize your OVS-DPDK deployment.

Note

You do not need to manually calculate these parameters if you use the derived_parameters.yaml workflow to generate these values automatically. See Overview of workflows and derived parameters

Note

Always pair CPU sibling threads (logical CPUs) together for the physical core when allocating CPU cores.

See Discovering your NUMA node topology to determine the CPU and NUMA nodes on your Compute nodes. You use this information to map CPU and other parameters to support the host, guest instance, and OVS-DPDK process needs.

7.4.1. CPU parameters

OVS-DPDK uses the following CPU partitioning parameters:

OvsPmdCoreList

Provides the CPU cores that are used for the DPDK poll mode drivers (PMD). Choose CPU cores that are associated with the local NUMA nodes of the DPDK interfaces. OvsPmdCoreList is used for the pmd-cpu-mask value in Open vSwitch.

  • Pair the sibling threads together.
  • Exclude all cores from the OvsDpdkCoreList
  • Avoid allocating the logical CPUs (both thread siblings) of the first physical core on both NUMA nodes as these should be used for the OvsDpdkCoreList parameter.
  • Performance depends on the number of physical cores allocated for this PMD Core list. On the NUMA node which is associated with DPDK NIC, allocate the required cores.
  • For NUMA nodes with a DPDK NIC:

    • Determine the number of physical cores required based on the performance requirement and include all the sibling threads (logical CPUs) for each physical core.
  • For NUMA nodes without DPDK NICs:

    • Allocate the sibling threads (logical CPUs) of one physical core (excluding the first physical core of the NUMA node).
Note

DPDK PMD threads must be reserved on both NUMA nodes even if a NUMA node does not have an associated DPDK NIC.

NovaVcpuPinSet

Sets cores for CPU pinning. The Compute node uses these cores for guest instances. NovaVcpuPinSet is used as the vcpu_pin_set value in the nova.conf file.

  • Exclude all cores from the OvsPmdCoreList and the OvsDpdkCoreList.
  • Include all remaining cores.
  • Pair the sibling threads together.
NovaComputeCpuSharedSet
Sets the cores to be used for emulator threads. This will define the value of the nova.conf parameter cpu_shared_set. The recommended value for this parameter matches the value set for OvsDpdkCoreList.
IsolCpusList

A set of CPU cores isolated from the host processes. This parameter is used as the isolated_cores value in the cpu-partitioning-variable.conf file for the tuned-profiles-cpu-partitioning component.

  • Match the list of cores in OvsPmdCoreList and NovaVcpuPinSet.
  • Pair the sibling threads together.
OvsDpdkCoreList

Provides CPU cores for non data path OVS-DPDK processes, such as handler and revalidator threads. This parameter has no impact on overall data path performance on multi-NUMA node hardware. This parameter is used for the dpdk-lcore-mask value in Open vSwitch, and these cores are shared with the host.

  • Allocate the first physical core (and sibling thread) from each NUMA node (even if the NUMA node has no associated DPDK NIC).
  • These cores must be mutually exclusive from the list of cores in OvsPmdCoreList and NovaVcpuPinSet.
DerivePciWhitelistEnabled

To reserve virtual functions (VF) for VMs, use the NovaPCIPassthrough parameter to create a list of VFs passed through to Nova. VMs excluded from the list remain available for the host.

Red Hat recommends that you change the DerivePciWhitelistEnabled value to false from the default of true, and then manually configure the list in the NovaPCIPassthrough parameter.

For each VF in the list, populate the address parameter with a regular expression that resolves to the address value.

The following is an example of the manual list creation process. If NIC partitioning is enabled in a device named eno2, list the PCI addresses of the VFs with the following command:

[heat-admin@compute-0 ~]$ ls -lh /sys/class/net/eno2/device/ | grep virtfn
lrwxrwxrwx. 1 root root    0 Apr 16 09:58 virtfn0 -> ../0000:18:06.0
lrwxrwxrwx. 1 root root    0 Apr 16 09:58 virtfn1 -> ../0000:18:06.1
lrwxrwxrwx. 1 root root    0 Apr 16 09:58 virtfn2 -> ../0000:18:06.2
lrwxrwxrwx. 1 root root    0 Apr 16 09:58 virtfn3 -> ../0000:18:06.3
lrwxrwxrwx. 1 root root    0 Apr 16 09:58 virtfn4 -> ../0000:18:06.4
lrwxrwxrwx. 1 root root    0 Apr 16 09:58 virtfn5 -> ../0000:18:06.5
lrwxrwxrwx. 1 root root    0 Apr 16 09:58 virtfn6 -> ../0000:18:06.6
lrwxrwxrwx. 1 root root    0 Apr 16 09:58 virtfn7 -> ../0000:18:06.7

In this case, the VFs 0, 4, and 6 are used by eno2 for NIC Partitioning. Manually configure NovaPCIPassthrough to include VFs 1-3, 5, and 7, and consequently exclude VFs 0,4, and 6, as in the following example:

NovaPCIPassthrough:
  - physical_network: "sriovnet2"
  address: {"domain": ".*", "bus": "18", "slot": "06", "function": "[1-3]"}
  - physical_network: "sriovnet2"
  address: {"domain": ".*", "bus": "18", "slot": "06", "function": "[5]"}
  - physical_network: "sriovnet2"
  address: {"domain": ".*", "bus": "18", "slot": "06", "function": "[7]"}

7.4.2. Memory parameters

OVS-DPDK uses the following memory parameters:

OvsDpdkMemoryChannels

Maps memory channels in the CPU per NUMA node. The OvsDpdkMemoryChannels parameter is used by Open vSwitch as the other_config:dpdk-extra=”-n <value>” value.

  • Use dmidecode -t memory or your hardware manual to determine the number of memory channels available.
  • Use ls /sys/devices/system/node/node* -d to determine the number of NUMA nodes.
  • Divide the number of memory channels available by the number of NUMA nodes.
NovaReservedHostMemory

Reserves memory in MB for tasks on the host. This value is used by the Compute node as the reserved_host_memory_mb value in nova.conf.

  • Use the static recommended value of 4096 MB.
OvsDpdkSocketMemory

Specifies the amount of memory in MB to pre-allocate from the hugepage pool, per NUMA node. This value is used by Open vSwitch as the other_config:dpdk-socket-mem value.

  • Provide as a comma-separated list. Calculate the OvsDpdkSocketMemory value from the MTU value of each NIC on the NUMA node.
  • For a NUMA node without a DPDK NIC, use the static recommendation of 1024 MB (1GB)
  • The following equation approximates the value for OvsDpdkSocketMemory:

    • MEMORY_REQD_PER_MTU = (ROUNDUP_PER_MTU + 800) * (4096 * 64) Bytes

      • 800 is the overhead value.
      • 4096 * 64 is the number of packets in the mempool.
  • Add the MEMORY_REQD_PER_MTU for each of the MTU values set on the NUMA node and add another 512 MB as buffer. Round the value up to a multiple of 1024.
Note

If the MTU size is not 1500, you might get a Failed to create memory pool error message in /var/log/messages. You can ignore this error message if it occurs at instance start up. To avoid this message, add the extra OvsDpdkSocketMemory amount for 1500 MTU onto your OvsDpdkSocketMemory calculation.

Sample Calculation - MTU 2000 and MTU 9000

DPDK NICs dpdk0 and dpdk1 are on the same NUMA node 0 and configured with MTUs 9000 and 2000 respectively. The sample calculation to derive the memory required is as follows:

  1. Round off the MTU values to the nearest 1024 bytes.

    The MTU value of 9000 becomes 9216 bytes.
    The MTU value of 2000 becomes 2048 bytes.
  2. Calculate the required memory for each MTU value based on these rounded byte values.

    Memory required for 9000 MTU = (9216 + 800) * (4096*64) = 2625634304
    Memory required for 2000 MTU = (2048 + 800) * (4096*64) = 746586112
  3. Calculate the combined total memory required, in bytes.

    2625634304 + 746586112 + 536870912 = 3909091328 bytes.

    This calculation represents (Memory required for MTU of 9000) + (Memory required for MTU of 2000) + (512 MB buffer).

  4. Convert the total memory required into MB.

    3909091328 / (1024*1024) = 3728 MB.
  5. Round this value up to the nearest 1024.

    3724 MB rounds up to 4096 MB.
  6. Use this value to set OvsDpdkSocketMemory.

        OvsDpdkSocketMemory: “4096,1024”

Sample Calculation - MTU 2000

DPDK NICs dpdk0 and dpdk1 are on the same NUMA node 0 and configured with MTUs 2000 and 2000 respectively. The sample calculation to derive the memory required is as follows:

  1. Round off the MTU values to the nearest 1024 bytes.

    The MTU value of 2000 becomes 2048 bytes.
  2. Calculate the required memory for each MTU value based on these rounded byte values.

    Memory required for 2000 MTU = (2048 + 800) * (4096*64) = 746586112
  3. Calculate the combined total memory required, in bytes.

    746586112 + 536870912 = 1283457024 bytes.

    This calculation represents (Memory required for MTU of 2000) + (512 MB buffer).

  4. Convert the total memory required into MB.

    1283457024 / (1024*1024) = 1224 MB.
  5. Round this value up to the nearest 1024.

    1224 MB rounds up to 2048 MB.
  6. Use this value to set OvsDpdkSocketMemory.

        OvsDpdkSocketMemory: “2048,1024”

7.4.3. Networking parameters

OvsDpdkDriverType
Sets the driver type used by DPDK. Use the default value of vfio-pci.
NeutronDatapathType
Datapath type for OVS bridges. DPDK uses the default value of netdev.
NeutronVhostuserSocketDir
Sets the vhost-user socket directory for OVS. Use /var/lib/vhost_sockets for vhost client mode.

7.4.4. Other parameters

NovaSchedulerDefaultFilters
Provides an ordered list of filters that the Compute node uses to find a matching Compute node for a requested guest instance.
VhostuserSocketGroup
Sets the vhost-user socket directory group. The default value is qemu. *VhostuserSocketGroup* should be set to hugetlbfs so that the ovs-vswitchd and qemu processes can access the shared hugepages and unix socket used to configure the virtio-net device. This value is role specific and should be applied to any role leveraging OVS-DPDK.
KernelArgs

Provides multiple kernel arguments to /etc/default/grub for the Compute node at boot time. Add the following based on your configuration:

  • hugepagesz: Sets the size of the huge pages on a CPU. This value can vary depending on the CPU hardware. Set to 1G for OVS-DPDK deployments (default_hugepagesz=1GB hugepagesz=1G). Check for the pdpe1gb CPU flag to ensure your CPU supports 1G.

    lshw -class processor | grep pdpe1gb
  • hugepages count: Sets the number of huge pages available. This value depends on the amount of host memory available. Use most of your available memory (excluding NovaReservedHostMemory). You must also configure the huge pages count value within the OpenStack flavor associated with your Compute nodes.
  • iommu: For Intel CPUs, add “intel_iommu=on iommu=pt”`
  • isolcpus: Sets the CPU cores to be tuned. This value matches IsolCpusList.

7.4.5. Instance extra specifications

Before deploying instances in an NFV environment, create a flavor that will utilize CPU pinning, emulator thread pinning, and huge pages.

hw:cpu_policy
Set the value of this parameter to dedicated, so that a guest will use pinned CPUs. Instances created from a flavor with this parameter set will have an effective overcommit ratio of 1:1. The default is shared.
hw:mem_page_size

Set the value of this parameter to a valid string of a specific value with standard suffix, for example, 4KB, 8MB, or 1GB. Use 1GB to match the hugepagesz boot parameter. The number of huge pages available for the virtual machines is the boot parameter minus the OvsDpdkSocketMemory. Other valid parameter values include the following:

  • small (default) - The smallest page size is used
  • large - Only use large page sizes. (2MB or 1GB on x86 architectures)
  • any - The compute driver may attempt large pages, but default to small if none available.
hw:emulator_threads_policy
Set the value of this parameter to share so that emulator threads are locked to CPUs that you’ve identified in the heat parameter, NovaComputeCpuSharedSet. If an emulator thread is running on a vCPU being used for the poll mode driver (PMD) or real-time processing, you can experience packet loss or missed deadlines.

7.5. Two NUMA node example OVS-DPDK deployment

This sample Compute node includes two NUMA nodes as follows:

  • NUMA 0 has cores 0-7. The sibling thread pairs are (0,1), (2,3), (4,5), and (6,7)
  • NUMA 1 has cores 8-15. The sibling thread pairs are (8,9), (10,11), (12,13), and (14,15).
  • Each NUMA node connects to a physical NIC (NIC1 on NUMA 0 and NIC2 on NUMA 1).
OpenStack NFV NUMA Nodes 453316 0717 ECE OVS DPDK Deployment
Note

Reserve the first physical cores (both thread pairs) on each NUMA node (0,1 and 8,9) for non data path DPDK processes (OvsDpdkCoreList).

This example also assumes a 1500 MTU configuration, so the OvsDpdkSocketMemory is the same for all use cases:

OvsDpdkSocketMemory: “1024,1024”

NIC 1 for DPDK, with one physical core for PMD

In this use case, you allocate one physical core on NUMA 0 for PMD. You must also allocate one physical core on NUMA 1, even though there is no DPDK enabled on the NIC for that NUMA node. The remaining cores (not reserved for OvsDpdkCoreList) are allocated for guest instances. The resulting parameter settings are:

OvsPmdCoreList: “2,3,10,11”
NovaVcpuPinSet: “4,5,6,7,12,13,14,15”

NIC 1 for DPDK, with two physical cores for PMD

In this use case, you allocate two physical cores on NUMA 0 for PMD. You must also allocate one physical core on NUMA 1, even though there is no DPDK enabled on the NIC for that NUMA node. The remaining cores (not reserved for OvsDpdkCoreList) are allocated for guest instances. The resulting parameter settings are:

OvsPmdCoreList: “2,3,4,5,10,11”
NovaVcpuPinSet: “6,7,12,13,14,15”

NIC 2 for DPDK, with one physical core for PMD

In this use case, you allocate one physical core on NUMA 1 for PMD. You must also allocate one physical core on NUMA 0, even though there is no DPDK enabled on the NIC for that NUMA node. The remaining cores (not reserved for OvsDpdkCoreList) are allocated for guest instances. The resulting parameter settings are:

OvsPmdCoreList: “2,3,10,11”
NovaVcpuPinSet: “4,5,6,7,12,13,14,15”

NIC 2 for DPDK, with two physical cores for PMD

In this use case, you allocate two physical cores on NUMA 1 for PMD. You must also allocate one physical core on NUMA 0, even though there is no DPDK enabled on the NIC for that NUMA node. The remaining cores (not reserved for OvsDpdkCoreList) are allocated for guest instances. The resulting parameter settings are:

OvsPmdCoreList: “2,3,10,11,12,13”
NovaVcpuPinSet: “4,5,6,7,14,15”

NIC 1 and NIC2 for DPDK, with two physical cores for PMD

In this use case, you allocate two physical cores on each NUMA node for PMD. The remaining cores (not reserved for OvsDpdkCoreList) are allocated for guest instances. The resulting parameter settings are:

OvsPmdCoreList: “2,3,4,5,10,11,12,13”
NovaVcpuPinSet: “6,7,14,15”
Note

Red Hat recommends using 1 physical core per NUMA node.

7.6. Topology of an NFV OVS-DPDK deployment

This example deployment shows an OVS-DPDK configuration and consists of two virtual network functions (VNFs) with two interfaces each:

  • The management interface, represented by mgt.
  • The data plane interface.

In the OVS-DPDK deployment, the VNFs operate with inbuilt DPDK that supports the physical interface. OVS-DPDK enables bonding at the vSwitch level. For improved performance in your OVS-DPDK deployment, it is recommended that you separate kernel and OVS-DPDK NICs. To separate the management (mgt) network, connected to the Base provider network for the virtual machine, ensure you have additional NICs. The Compute node consists of two regular NICs for the Red Hat OpenStack Platform API management that can be reused by the Ceph API but cannot be shared with any OpenStack project.

NFV OVS-DPDK deployment

NFV OVS-DPDK topology

The following image shows the topology for OVS_DPDK for the NFV use case. It consists of Compute and Controller nodes with 1 or 10 Gbps NICs, and the Director node.

NFV OVS-DPDK Topology