Netdev 1.1 - BoFs http://www.netdevconf.org/1.1/tags/bofs en BoF: "Describing datapath processing from userspace policy" (Brenden Blanco, John Fastabend) http://www.netdevconf.org/1.1/bof-describing-datapath-processing-userspace-policy-brenden-blanco-john-fastabend <div class="field field-name-body field-type-text-with-summary field-label-hidden"><div class="field-items"><div class="field-item even" property="content:encoded"><strong>Description</strong>: <p>There is an increasing demand for languages and tools to describe datapath processing from userspace policy. This includes packet parsing, mangling, lookups and forwarding actions. This BoF will collect the experience of ongoing efforts from different fronts to resolve this problem in the Linux networking area.</p></div></div></div><div class="field field-name-field-tags field-type-taxonomy-term-reference field-label-above clearfix"><h3 class="field-label">Tags: </h3><ul class="links"><li class="taxonomy-term-reference-0" rel="dc:subject"><a href="/netdev/tags/bofs" typeof="skos:Concept" property="rdfs:label skos:prefLabel" datatype="">BoFs</a></li></ul></div> Sat, 06 Feb 2016 11:21:18 +0000 admin 82 at http://www.netdevconf.org/1.1 BoF: "Network Performance BoF" (Jesper Dangaard Brouer) http://www.netdevconf.org/1.1/bof-network-performance-bof-jesper-dangaard-brouer <div class="field field-name-body field-type-text-with-summary field-label-hidden"><div class="field-items"><div class="field-item even" property="content:encoded"><strong>Description</strong>: <p>This BoF is not about the amazing performance improvements we have completed over the last year. This BoF is about existing bottlenecks observed in the kernel network-stack, and how to tackle these.</p> <p>It is a forum where we discuss how to address and tackle these existing bottlenecks.</p> <p>We call for speakers, wanting to share their observations and who seek input from the community and the kernel network developers present at this conference.</p> <p>The format; for each subject area, the speaker must limit the number of slides (2-5) describing the bottleneck and use-case it occurs in. We do encourage speakers to propose several possible ways they imagine to tackle the bottleneck, but leaves it up to the audience to discuss which path is the most fruitful. We do encourage new bold and crazy ideas to surface.</p> <p>This session hope to see kernel developers with deep code level knowledge of the Linux kernel network stack.</p> </div></div></div><div class="field field-name-field-tags field-type-taxonomy-term-reference field-label-above clearfix"><h3 class="field-label">Tags: </h3><ul class="links"><li class="taxonomy-term-reference-0" rel="dc:subject"><a href="/netdev/tags/bofs" typeof="skos:Concept" property="rdfs:label skos:prefLabel" datatype="">BoFs</a></li></ul></div> Wed, 06 Jan 2016 10:48:25 +0000 admin 58 at http://www.netdevconf.org/1.1 BoF: "switchdev" (Roopa Prabhu & Shrijeet Mukherjee) http://www.netdevconf.org/1.1/bof-switchdev-roopa-prabhu-shrijeet-mukherjee <div class="field field-name-body field-type-text-with-summary field-label-hidden"><div class="field-items"><div class="field-item even" property="content:encoded"><strong>Description</strong>: <p>The switchdev project in the Linux Kernel has come a long way in preserving the Linux networking API across devices with differing capabilities. It has provided a path to unify networking configuration across all devices running Linux (Servers, small wireless devices and data center switches). Over the past year (thanks to its maintainers and the community!) the project has seen many networking drivers (notably dsa) move to this new infrastructure and it also recently received a 100G switch driver user!</p> <p>The purpose of this BOF is to discuss status/open-issues and HW offload API's and support for newer networking features:</p> <ul> <li>Offloading ACL's to HW: choices and the offload API.</li> <li>HW resource accounting.</li> <li>Offload error handling policy.</li> <li>New offload API's for vxlan, IGMP snooping and policy routing.</li> </ul> </div></div></div><div class="field field-name-field-tags field-type-taxonomy-term-reference field-label-above clearfix"><h3 class="field-label">Tags: </h3><ul class="links"><li class="taxonomy-term-reference-0" rel="dc:subject"><a href="/netdev/tags/bofs" typeof="skos:Concept" property="rdfs:label skos:prefLabel" datatype="">BoFs</a></li></ul></div> Sun, 03 Jan 2016 10:33:08 +0000 admin 54 at http://www.netdevconf.org/1.1 BoF: "IPsec performance" (Steffen Klassert) http://www.netdevconf.org/1.1/bof-ipsec-performance-steffen-klassert <div class="field field-name-body field-type-text-with-summary field-label-hidden"><div class="field-items"><div class="field-item even" property="content:encoded"><strong>Description</strong>: <p>IPsec suffers from poor performance compared to non IPsec protocols. This is mostly because the needed crypto operations are cpu intensive, but also the IPsec networking path is not well optimized.</p> <p>This BoF is to discuss possible improvements of the IPsec networking path.</p> <p>Here we want to consider pure software improvements as well as hardware support for IPsec.</p> <p>Possible improvements could be:</p> <ul> <li>Avoid copying most of the data frames before doing crypto.</li> <li>Adding a software GRO/GSO codepath for IPsec.</li> <li>Support IPsec offloading to NICs.</li> <li>Change xfrm_policy_lock from rwlock to rcu.</li> </ul> <p>The following people should attend:</p> <ul> <li>Network developers with IPsec background.</li> <li>Network driver developers whose hardware supports IPsec.</li> <li>IPsec users.</li> </ul></div></div></div><div class="field field-name-field-tags field-type-taxonomy-term-reference field-label-above clearfix"><h3 class="field-label">Tags: </h3><ul class="links"><li class="taxonomy-term-reference-0" rel="dc:subject"><a href="/netdev/tags/bofs" typeof="skos:Concept" property="rdfs:label skos:prefLabel" datatype="">BoFs</a></li></ul></div> Sat, 26 Dec 2015 11:06:31 +0000 admin 46 at http://www.netdevconf.org/1.1 BoF: "Unlocking SR-IOV in Linux" (John Fastabend) http://www.netdevconf.org/1.1/bof-unlocking-sr-iov-linux-john-fastabend <div class="field field-name-body field-type-text-with-summary field-label-hidden"><div class="field-items"><div class="field-item even" property="content:encoded"> <strong>Description</strong>: <p>The Linux kernel does not fully utilize existing hardware capabilities provided by SR-IOV. As a result user-space consumers (such as libvirt) end up only supporting the most basic of features. The result is serious users of SR-IOV end up depending on proprietary implementations outside of Linux.</p> <p>The purpose of this BOF is to get SR-IOV on Linux unstuck from the current stall.</p> <p>Any of the following are highly encouraged to attend:</p> <ul> <li>Driver developers.</li> <li>Network developers.</li> <li>User space developers consuming SR-IOV.</li> <li>Users interested in consuming SR-IOV.</li> </ul> <p>A tentative agenda (we would like to solicit topics as well):</p> <ul> <li>Layer 3 support for virtual functions.</li> <li>Wildcard lookups for ACLs and forwarding.</li> <li>Tunnel support, including encap/decap.</li> <li>Virtual function live migration.</li> <li>PF promiscuous support.</li> <li>Monitoring support.</li> <li>Integration with OVS, bridge, 'tc', and netfilter.</li> </ul> <p>Note: Alex Duyck's <a href="http://www.netdevconf.org/1.1/tutorial-using-sr-iov-openstack-alexander-duyck">tutorial</a> is a good introduction to SR-IOV and both beginners and experienced netdevers are encouraged to attend that tutorial before showing up for the BOF.</p></div></div></div><div class="field field-name-field-tags field-type-taxonomy-term-reference field-label-above clearfix"><h3 class="field-label">Tags: </h3><ul class="links"><li class="taxonomy-term-reference-0" rel="dc:subject"><a href="/netdev/tags/bofs" typeof="skos:Concept" property="rdfs:label skos:prefLabel" datatype="">BoFs</a></li></ul></div> Wed, 09 Dec 2015 17:15:17 +0000 admin 40 at http://www.netdevconf.org/1.1