Hmmm.... there are a couple of things to consider in your question.... FRR is the routing protocol, so the types of performance numbers that typically make sense in that context are convergence time given a route change event (i.e. join a network, topology change, etc). Packet throughput is well outside of the FRR domain, which leads me to the second observation... Packet throughput on an X86 platform is based on a TON of factors... the kernel version, kernel tuning, NIC type, NIC drivers, NIC firmware (if required), other invoked SW/networking features (iptables, tc, etc). With all of that said, I know of customers running FRR on X86 systems and getting 40G routing performance. cheers, JR On Wed, Jul 5, 2017 at 4:44 AM, Richard Barger <rbarger@ecdata.com> wrote:

Hello,

I would like to ask if there are any performance numbers for 10GbE packet throughput with the latest release of FFR and BGP.

Such as a RFC-2544 throughput test result….

We are using Intel X710 10GbE NIC cards and new INTEL XEON E5V3-4 based servers.

We have used QUAGGA prior to build a BGP Network Latency Emulator on 10/100/1000 XEON E5-1620V2 system.

The packet throughput was not great.

So, we are very interested to learn of the performance enhancements of FRR over prior Quagga from three years ago.

We are wanting to understand how to obtain the fastest possible Packet throughput.

Sincerely,

Richard Barger, President

*East Coast Datacom, Inc.*

245 Gus Hipp Blvd., STE 3

Rockledge, FL 32955 USA

Tel: (321) 637-9922

Fax: (321) 637-9980

email: rbarger@ecdata.com

*Web: www.ecdata.com <http://www.ecdata.com>*

[image: LOGO ecdata]

_______________________________________________ dev mailing list dev@lists.frrouting.org https://lists.frrouting.org/listinfo/dev