_______________________________________________Hi Sim,
Depends on your performance requirements. Are you trying to push 10G+/14M PPS? If so, you probably want to look into VPP. VPP is a kernel bypass dataplane implementation that is supposed to drastically improve forwarding performance over the Linux kernel. Decent info here: https://wiki.fd.io/view/VPP/What_is_VPP%3F
Matt Love
Sr. Network Engineer
Omnificent Systems Inc. | osCloud | AS14525
From: Sim <simvirus@gmail.com>
Date: Friday, February 22, 2019 at 6:24 AM
To: Matt Love <matt@oscloud.io>
Cc: "frog@lists.frrouting.org" <frog@lists.frrouting.org>
Subject: Re: [FROG] Hardware Requirements
Hello Matt,
thanks for sharing your experience.
I'm using Ubuntu 18.04 LTS + FRR + TC QoS (in VMware environment with two vNIC)
First NIC is for "BORDER" and the secondo NIC is for "INTERIOR".
No-NAT, No-Firewall or No-other... only BGP/Routing.
Well... I do not know VPP.
Could you explain me if it could be interesting in my infrastructure/config?
Regards
Sim
Il giorno gio 21 feb 2019 alle ore 16:08 Matt Love <matt@oscloud.io> ha scritto:
Hello,
I’m currently in exactly the same boat. I spent a good bit of time getting VPP and FRRouting married on a SuperMicro chassis with 4x Intel X722 NICs. I have a loose collection of notes/instructions here:
Ultimately we haven’t moved with this in production due to lack of faith in performance. I haven’t been able to prove that it’ll do 14M PPS to achieve line rate 10 Gbps, mainly because I don’t have a testing environment.
But, in the mean time, feel free to peruse my docs, they should help you build a semi-reliable VPP/FRR setup.
Cheers
Matt Love
Sr. Network Engineer
Omnificent Systems Inc. | osCloud | AS14525
⌁ Sent from Mobile
On Feb 21, 2019, at 05:34, Steuer Heribert <steuer@patronas.com> wrote:Hello Donald,
thanks for your reply. I think it would be quite useful for users to have an overview of proven hardware setups. Control-plane is finally just one part of the game. We finally want to replace some
old Cisco edge routers with a handful of BGP peers (all full table) for IPv4 and IPv6. Bandwidth per port does not exceed 10gbps, but to be honest I do not know how FRR will perform without any optimized data-plane in this setup.
Any experience?
Cheers,
Heri
Von: Donald Sharp <sharpd@cumulusnetworks.com>
Datum: Donnerstag, 21. Februar 2019 um 13:23
An: Steuer Heribert <steuer@patronas.com>
Cc: "frog@lists.frrouting.org" <frog@lists.frrouting.org>
Betreff: Re: [FROG] Hardware Requirements
Steuer -
This is not something that we've put together at this point in time. FRR works quite nicely on all modern processors but you seem more concerned about data plane forwarding, which while understandable, is not something we have a bunch of control over. As I understand it using something like XDP or DPDK works quite nicely as accelerators for packet processing, it is just not something I have experience with.
donald
On Thu, Feb 21, 2019 at 7:20 AM Steuer Heribert <steuer@patronas.com> wrote:
Hello,
for using FRR with VPP & underlying DPDK, is there any “working hardware” list which goes beyond “use Intel NICs”? I am aware of some white box lists provided by NOS vendors but was not able to find anything about native FRR. I´ve found some posts of cloudscale.ch where they mentioned appliances from a vendor named Lanner (which I never heard of before). Its not about advertising vendors but more kind of a “known-to-work-well” list.
In addition it would be nice to chat on Slack, please invite 😊
Thanks,
Heri
_______________________________________________
frog mailing list
frog@lists.frrouting.org
https://lists.frrouting.org/listinfo/frog_______________________________________________
frog mailing list
frog@lists.frrouting.org
https://lists.frrouting.org/listinfo/frog_______________________________________________
frog mailing list
frog@lists.frrouting.org
https://lists.frrouting.org/listinfo/frog
frog mailing list
frog@lists.frrouting.org
https://lists.frrouting.org/listinfo/frog