Mellanox MCX516A-BDAT does it support XDP and can we direct connect it to an adapter with QSFP+

Infrastructure & Networking Adapters and Cables
1

Hello,

we’re thinking of arranging our 40G packet sniffers with Mellanox MCX516A-BDAT. But we have some considerations regarding:

  1. Does it support XDP ? Does all the NICs ConnectX-5 and ConnectX-4 support XDP
  2. EoL of the product ?
  3. Can we plug in a QSFP+ transceiver module so we can connect on the other end to another 40G QSFP+ (direct connection)
  4. If the NIC is 100G by default (MCX516A-CDAT for example**)** can we breakout to a 40G (like scenario number 3)

Thank you very much.

2

Hello Alexander,

Thank you for posting your inquiry on the NVIDIA Networking Community.

Based on the information provided:

  1. From ConnectX-4 and above we support XDP → https://docs.mellanox.com/display/MLNXENv531001/Release+Notes+Change+Log+History (Search for XDP). The following page provides an example on how-to run on ConnectX-5 → https://docs.mellanox.com/display/MLNXENv531001/Release+Notes+Change+Log+History

  2. EOL announcements can be found through the following link → https://mellanox.com/support/eol

  3. Yes, You can use a validated/supported cable/transceiver to connect back-2-back between to adapters, without the use of a switch. See the following link for the supported cable for the ConnectX-5 → https://docs.mellanox.com/display/ConnectX5Firmwarev16311014/Firmware+Compatible+Products#FirmwareCompatibleProducts-ValidatedandSupported40GbECables

You can configure the required speed through ‘ethtool’ which is available for many Linux distro’s

  1. No, the break-out functionality is only available for our switches. Our adapters do not have the capability to break-out their ports

ConnectX-4

Thank you and regards,

~NVIDIA Networking Technical Support

3

Thanks Martijn for your detailed answer.

Can I please be answered one more thing regarding XDP support.

What level of XDP support there is?

Is it Native (driver) XDP: The kernel executes the program from the earliest possible point during packet reception. At this moment, the kernel did not parse the packet and, therefore, no metadata provided by the kernel is available. This mode requires that the network interface driver supports XDP but not all drivers support this native mode.

Or is it…

Offloaded XDP: The kernel executes the XDP program on the network interface instead of on the host CPU. Note that this requires specific hardware, and only certain eBPF features are available in this mode.?

(as described in - XDP section here

Chapter 55. Understanding the eBPF networking features in RHEL Red Hat Enterprise Linux 8 | Red Hat Customer Portal)

4

Hello,

can you kindly advise. We’re waiting for the answer so we can make an informative decision here.

Thank you.