Hi, Gleaned this thread on nanog mailer archive [1] mentioning fd.io :) Not a "thought leader", but I thought I respond with some fd.io data points, hopefully relevant to the thread’s subject. Note that most of them focus on data plane functionality, performance and technology level comparison: - IPv4 routing (large FIB) at Terabit speed, a video clip (low-budget) showing what difference modern Xeon processors (Skylake) can make to SW data plane if coupled with something super-fast, produced by LF FD.io project for Xeon Skylake launch on 11th July 2017 (a year ago). [2] - SW data plane benchmarking, technical paper [3][4]. - Test and performance report for last fd.io release [5]. - Continuous trending of fd.io vpp master branch [6]. And although network control plane is not covered (e.g. BGP), many data plane hooks required are there incl. large FIBs, VRFs, hierarchical FIB for IP fast convergence, multicast, etc. HTH -Maciek [1] https://mailman.nanog.org/pipermail/nanog/2018-June/095897.html [2] "FD.io: A Universal Terabit Network Dataplane", https://www.youtube.com/watch?v=aLJ0XLeV3V4. [3] https://fd.io/wp-content/uploads/sites/34/2018/01/performance_analysis_sw_da... [4] https://wiki.fd.io/images/3/31/Benchmarking-sw-data-planes-Dec5_2017.pdf [5] FD.io CSIT-CPL rls1804, https://docs.fd.io/csit/rls1804/report/. [6] FD.io CSIT-CPL continuous VPP performance trending: IPv4, https://docs.fd.io/csit/master/trending/new/trending/ip4.html as of next week above link will be migrated to: https://docs.fd.io/csit/master/trending/trending/ip4.html IPv6, https://docs.fd.io/csit/master/trending/new/trending/ip6.html as of next week above link will be migrated to: https://docs.fd.io/csit/master/trending/trending/ip6.html
From: Marcus Leske <marcusleskex@gmail.com> Date: June 14, 2018 at 07:18:16 PDT To: nanog@nanog.org Subject: fd.io vs cumulus vs snabb vs OVS vs OpenNSL
Hi
Any thought leader on the list to shed some light to what is happening in the world of open networking ? OVS vs OpenNSL vs Cumulus vs fd.io vs Snabb vs a lot of stuff :)
Where is this going ? What are the obvious pros and cons of each when it comes to scale and feature velocity ?
https://www.reddit.com/r/networking/comments/8r0afq/is_their_any_truth_to_th...
Danke