- ovscon.site akaris blog which also has a summary of ovscon blog posts to read to do next terminal presentations use asciinema mapping networks with skydive
- openstack redhat doc
- note that the rpm is not available for RHOSP 15, only for 14. This was a tech preview
- OSP skydive presentation
- spec dragonflow making use of skydive make sense of MPLS/VPLS
- ovs orbit podcast https://ovsorbit.org/
- switching from C to rust language for better mem mgmt
- not part of the upstream branch for now, still in testing/beta
- Physical to Virtual back to Physical topology (PVP)
- Automated Open vSwitch PVP testing (setup trex, testpmd and run the PVP scripts) https://github.com/chaudron/ovs_perf
- Tuple space explosion, MegaFlowCache flow exploit
- generic TSE Attack impacts
- MFC Guard (MFCg) "Mitigation technique with a cache management scheme called MFC Guard (MFCg). It dynamically monitors the number of entries in the MFC and removes less important ones to reduce the performance overhead of the TSS algorithm. It is worth noting that we have observed negligible impact on the overall packet processing performance during the monitoring process itself (i.e., executing ovs-dpctl dump-flows in, say, each second)."
- vmware driven ovs based kubes nw plugin antrea
- Where RH is driving ovn-kubernetes plugin instead
- kubes network plugins
- presenting INTEL flavour of kubes ovs plugin (does support dpdk/sriov too)
- Multus
- it apparently already semi-works with standard kuryr integration
- Intel
- cpu instruction implementation (scalar and avx-512)
- suggestion of new way to compile ovs for specific cpu arch
- Intel
- project to fill the gaps from ovn-kubernetes plugging, mainly for VNFs
- ovn for nfv k8s (ovn4nfvk8s)
- This project works with Multus to provide a CNI to attach multiple OVN interfaces into workloads and supports CRD controllers for dynamic networks and provider networks and for route management.
- basically bridging ovn-kubernetes plugin with ovn4nfvk8s to use multus
- Broadcom
- rte_flow_validate not used by dpdk, need to optimize capability discovery
- dynamic rebalance, apparently needed for capacity mgmt (need to read up)
- rte_flow on Intel nics 700 hw offload
- allows standard flow is accelerated at MF Extraction, Exact Match Cache (EMC) and DPCLS (megaflow cache) lookups
PKT-IN --> MF-EXTRACT --> LOOKUP --> ACTION --> PKT-OUT
More to read:
- pipeline change
- design and implementation of the datapath classifier
- ovs lookup presentation 2016
- OvS-DPDK Datapath Classifier – Part 2
- Netronome presentation
- Virtual Switch Acceleration with OVS-TC
- Multiprotocol Label Switching
- vDPA netdev is designed to support both SW and HW
- HW mode will be used to configure vDPA capable devices. SW acceleration is used to leverage SRIOV offloads to virtio guests by relaying packets between VF and virtio devices. Add the SW relay forwarding logic as a pre-step for adding dpdkvdpa port with no functional change.
- Packet classifier for Linux kernel traffic classification (TC) subsystem ovs-tc
- operators view of tc-hw-offload on commodity x86
- ovs tc hw offload sucks netdev-tc-offload
- very llittle documentation
- hard to tell when the buffer gets overrun and impacts performance
- PCIv3 on an x8 lanes slot hit a PCI bottleneck (apparently there's a flag to use to avoid hitting this)
- prb node brd will flood all the switch ports, multicast_grp sequentially executes the pipelines
- Ubuntu marketing presentation
- ovn-openflow-probe reconnection 5 secs, can cause cpu consumption issues, resolution is to increase the probe increment
- dhcp reply time if loadedan cause snowball issue, where VIF will resend and flood
- continuous arp, update to ovn 2.12 (if lflow)run takes 10s > 100% cpu usage)
- need better debugging tools, ovn-trace and ovn-detrace arent clear enough (something like Skydive would be nice)
- ebay
- virtio in hardware
- virtio hw offload (in progress)
- vDPA
- nuage/mellanox/redhat
- Facebook product
- goal to bring LTE everywhere cheap