You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 2 Next »


Unable to render {include} The included page could not be found.

This section describes the hardware and software requirements for the SBC SWe on an OpenStack platform.

OpenStack Requirements

The SBC SWe supports the following OpenStack environments:

  • Newton with RHOSP 10 and RHEL 7.4
  • Queens with RHOSP 13 and RHEL 7.5
  • Train with RHOSP 16.1.1 and RHEL 8.2

Server Hardware Requirements

ConfigurationRequirement
Processor

Intel Xeon processors (Nehalem micro-architecture or above) with 6 cores and above (processors should support hyper-threading).

Note

Ribbon recommends Westmere (or newer) processors for better SRTP performance. These processors have the AES-NI instruction set for performing cryptographic operations in hardware. 

 RAMMinimum 24 GiB
 Hard DiskMinimum 100 GB
Network Interface Cards (NICs)

Minimum 4 NICs.

The following are supported for configuring as SR-IOV and DirectPath I/O pass-through devices:

  • Intel I350
  • Intel x540
  • Intel x550
  • Intel x710
  • Intel 82599 Ethernet adapters
  • Mellanox Connect - 4x
  • Mellanox Connect - 5x
  • QLogic 536LR-T

Note

Make sure NICs have multi-queue support which enhances network performance by allowing RX and TX queues to scale with the number of CPUs on multi-processor systems.

Note

The PKT ports must be 10 Gbps SR-IOV enabled ports.

Note

For packet port redundancy:

SRIOV

  • A minimum of 2 NICs are required.

DIO

  • A minimum of 4 NICs are required.

Note

To configure VLAN on SRIOV and PCI Passthrough Ethernet interfaces, disable the Data Center Bridging (DCB) on the switch connected to the interfaces.


S-SBC SWe Requirements

The system hosting the SBC SWe must meet the following requirements to achieve the performance targets listed: 

S-SBC SWe Requirements
for 1000 CPS/120K Signaling Sessions 		Notes

32 vCPUs

Due to the workload characteristics, allocate 20 physical cores with two hyper-threaded CPUs from each core to the SBC.

128 GiB RAM

Must be Huge Page memory. The minimum page size is 2048 KiB, but 1048576 is recommended.

100 GB Disk

None

4 vNICs/6 vNICs

Attach MGT0 port to the Management VirtIO Tenant network.

HA port has to be on IPv4 VirtIO Tenant network.

Attach PKT0 and PKT1 ports to SR-IOV and Provider network.

You must have 6 vNICs to enable PKT port redundancy. For more information, refer to the SBC SWe Features Guide.

Note

All NIC ports must come from the NUMA node 0. The S-SBC SWe instance is hosted on dual-socket physical server with 10 physical cores coming from each NUMA node.


M-SBC SWe Requirements

M-SBC SWe Requirements
for 40K Media Sessions		Notes

16 vCPUs

Due to the workload characteristics, allocate 10 physical cores with two hyper-threaded CPUs from each core and from single NUMA node to the SBC.

32 GiB RAM

Must be Huge Page memory. The minimum page size is 2048 KiB, but 1048576 is recommended.

100 GB Disk

None

4 vNICs/ 6 vNICs

Attach MGT0 port to the Management VirtIO Tenant network.

HA port has to be on IPv4 VirtIO Tenant network.

Attach PKT0 and PKT1 ports to SR-IOV and Provider network.

You must have 6 vNICs to enable PKT port redundancy. For more information, refer to the SBC SWe Features Guide.

Note

All NIC ports must come from the same NUMA node from which the M-SBC SWe instance is hosted.

OAM Node Requirements

OAM Node (minimum)Notes

2 vCPUs

None

16 GiB RAM

None

80 GB Disk

None

2 vNICs

None

  

I-SBC SWe Requirements

I-SBC SWe RequirementsNotes

20 vCPUs


32 GiB RAM

Must be Huge Page memory. The minimum page size is 2048 KiB, but 1048576 is recommended.

100 GB Disk

None

4 vNICs/ 6 vNICs

Attach MGT0 port to the Management VirtIO Tenant network.

HA port has to be on IPv4 VirtIO Tenant network.

Attach PKT0 and PKT1 ports to SR-IOV and Provider network.

You must have 6 vNICs to enable PKT port redundancy. For more information, refer to the SBC SWe Features Guide.


Note

For deployments that require it, you can instantiate the SBC SWe in smaller-sized configurations that use limited memory and vCPU resources. However, the limited resources place some restrictions on capacity and capabilities. Refer to Small SBC SWe Deployment Characteristics and Small SBC SWe Configuration Performance Metrics for more information.

Recommended Host Settings: Setting disk cache mode to "writethrough"

In order to prevent the SBC from crashing when the compute host power cycle is reset, it is recommended to set the cache mode for the disk type "file" to "writethrough". Execute the following steps on the compute host:

  1. Retrieve the current disk cache modes in the nova configuration on the compute host by executing the following command:

    Command
    crudini --get /var/lib/config-data/puppet-generated/nova_libvirt/etc/nova/nova.conf libvirt disk_cachemodes
    Sample Output
    file=none,block=writeback,network=writeback
    Note

    Note that "file=none" in the sample output above. Continue to Step 2 to set "file=writethrough".


  2. Set the disk type "file" to "writethrough" by executing the following command:

    Command
    crudini --set /var/lib/config-data/puppet-generated/nova_libvirt/etc/nova/nova.conf libvirt disk_cachemodes 'file=writethrough,block=writeback,network=writeback'
    Note

    The cache mode for the two other disk types "block" and "network" are determined by the customer. The available cache modes are "writethrough", "writeback", and "none". 

  3. Restart all nova containers by executing the following command:

    Command
    docker restart $(docker ps|grep nova|sed -r 's/^([^ ]+).*/\1/')
    Note

    The container restart procedure may vary depending on the container management tool. When in doubt, reboot the compute host.

  4. Verify in the nova configuration that the cache mode "file" is set to "writethrough" by executing the following command:

    Command
    crudini --get /var/lib/config-data/puppet-generated/nova_libvirt/etc/nova/nova.conf libvirt disk_cachemodes
    Sample Output
    file=writethrough,block=writeback,network=writeback

  5. If any VMs/instances exist on the compute host, verify in libvertd that "cache='writethrough'" for the disk type "file" by executing the following command:

    Command
    virsh dumpxml <instance_name> | grep -A 1 "disk type='file'"
    Note

    In the command above, replace <instance_name> with the actual name of the instance.

    Sample Output
    <disk type='file' device='disk'>
      <driver name='qemu' type='qcow2' cache='writethrough'/>


  • No labels