Got it

Low Bandwidth of IB Port Links on a Dell R6525 Compute Node

127 0 0 0 0

Problem Information

Basic information

Item

Description

Storage type

Distributed storage

Product version

OceanStor 100D 8.0.3; OceanStor Pacific series 8.1.0

Problem type

Performance maintenance

Keyword

DPC; Dell R6525 servers; CX-5; perftest; IB port link bandwidth


Problem Symptom

Dell R6525 servers (AMD CPUs) function as compute nodes. Each server has one Mellanox CX-5 100 Gbit/s IB NIC (model: MCX555A-ECAT). Compute nodes and distributed storage nodes are interconnected through IB switches.

Use ib_send_bw in the perftest suite of the MLNX_OFED NIC driver to test the maximum bandwidth of physical links of the IB network ports on two compute nodes.

In a scenario where the I/O size is 4 KB or 1 MB and the number of QPs (RDMA connections) is 1 or 10, the maximum bandwidth of a physical link is far lower than 100 Gbit/s.

Node Type

Number of QP Connections

I/O Block Size

Link Bandwidth (Gbit/s)

Compute nodes 01 and 02

1

4 KB

65

1 MB

95

10

4 KB

21

1 MB

25


Problem Diagnosis

Check the MLNX_OFED driver version, DELL server mainboard BIOS parameter settings, and IB NIC firmware parameter settings.

Solution

  1. X2APIC Mode is set to Disabled for the mainboard BIOS of the Dell R6525 server.

    2_en-us_image_0303626471.png

  2. Restart the compute node.

  3. Upgrade the IB NIC (Mellanox MCX555A-ECAT) firmware to the latest version (for example, 16.28.2006) provided by the Mellanox official website.

    #mst start

    #mst status

    #flint -d /dev/mst/mt4119_pciconf0 -i xxx.bin b

    2_en-us_image_0303626472.png

  4. Restart the compute node.

  5. Modify the configuration of the IB NIC firmware as follows:

    #mst start

    #mst status

    #mlxconfig -y -d /dev/mst/mt4119_pciconf0 q |grep PCI_WR

    2_en-us_image_0303626473.jpg

    #mlxconfig -y -d /dev/mst/mt4119_pciconf0 s PCI_WR_ORDERING=1

  6. Restart the compute node.

  7. Check whether the configuration is successful.

    #mst start

    #mst status

    #mlxconfig -y -d /dev/mst/mt4119_pciconf0 q |grep PCI_WR

    2_en-us_image_0303626474.jpg

Check After Recovery

  1. Multiple compute nodes can ping each other to ensure that the IB network is reachable.

  2. Run the ib_send_bw command to test the IB channel link again.

Suggestion and Summary

N/A

Applicable Versions

OceanStor 100D 8.0.3; OceanStor Pacific series 8.1.0

Comment

You need to log in to comment to the post Login | Register

Notice: To protect the legitimate rights and interests of you, the community, and third parties, do not release content that may bring legal risks to all parties, including but are not limited to the following:
  • Politically sensitive content
  • Content concerning pornography, gambling, and drug abuse
  • Content that may disclose or infringe upon others ' commercial secrets, intellectual properties, including trade marks, copyrights, and patents, and personal privacy
Do not share your account and password with others. All operations performed using your account will be regarded as your own actions and all consequences arising therefrom will be borne by you. For details, see " User Agreement."

My Followers

Login and enjoy all the member benefits

Login

Block
Are you sure to block this user?
Users on your blacklist cannot comment on your post,cannot mention you, cannot send you private messages.
Reminder
Please bind your phone number to obtain invitation bonus.