DPDK and FD.io Mini Summit
Tuesday, June 13
9:00 - 17:00
The Data Plane Development Kit (DPDK) is a set of drivers and software libraries for accelerating packet processing workloads on general purpose processors. It’s an open source, vendor-neutral project that runs on multiple different CPU platforms (x86, IA32, Power 8, ARMv7/v8) and supports NICs from a wide variety of vendors. It’s used in many different industry segments, including wireless core and access, wireline, cable, network appliances, cloud, and enterprise. DPDK is one of the foundation stones for NFV, providing the packet processing performance required to make NFV a reality.
Join us as the DPDK mini summit to hear about the latest developments in the project from community experts. Topics covered will include the DPDK roadmap, DPDK support in containers, the extension of the DPDK framework to support new accelerator types, and DPDK support for Service Assurance requirements defined by OPNFV.
|9:00 - 9:40||Introduction & DPDK Roadmap||Tim O'Driscoll (Intel)||This session will introduce the DPDK/FD.io mini summit, and describe the objectives and agenda for the day. It will also provide an overview of the current status of the DPDK project, describe some of the key recent developments, and will outline future plans.|
|9:40 - 10:20||Data Plane Virtualization Enhancements||Cunming (Steve) Liang (Intel)||This presentation will describe the main virtualization enhancements that have been included in recent DPDK releases, and will also cover future plans for further improvements.|
|10:20 - 11:00||Putting OVS-DPDK into Massive Production, with 0.0% Frame Loss||Franck Baudin & Christian Trautman (Red Hat)||Putting a new piece of software in Telco production environment, and manage its lifecycle requires a rock-solid CI. VSPerf, developed and deployed within OPNFV has also be internalized within Red Hat CI: it provides automated performance measurement of the bare-metal platform, testpmd as a vswitch, and OVS-DPDK for various configuration. In this session we will deep dive into the various test scenarios, the tests results, and see how they can be used by the various communities: DPDK developers, working on virtio/vhost but also on native PMD drivers, OVS community, FD.io community, OpenStack community, OPNFV community.
During this session we will also deep dive on the recipe, tested by VSPerf, that ensure zero packet loss with OVS-DPDK.
|11:00 - 11:15||Break|
|11:15 - 11:55||Cloud Monitoring Framework based on DPDK||Waterman Cao & Wei Yue (Huawei)||With the booming of container technology, it brings obvious advantages for cloud: simple and faster deployment, portability and lightweight cost. SDN/NFV/DPDK increases cloud capability and intelligence for more functionality, performance boost and higher resource utilization, etc. But it brings significant new challenges as well. One big concern is cloud monitoring is still relatively dumb. For example, we lack the ability to live detect sub-optimal symptoms, burst events, etc., which often leaves DevOps in blind. In order to meet these kind of challenges, we need to be able to monitor end-to-end traffic path in complexity cloud network environment. We designed a generic DSL-driven distributed monitor framework for this purpose. This framework deploys monitoring agents in various data-paths, including high-speed in-path DPDK monitor agent. It allows dynamic monitoring events injection and has the capability to monitor proactively with finer granularity to allow quick bottleneck detection and sub-optimal symptom alert.|
|12:55 - 12:30||Virtio-User as an Exceptional Path - New Path to kernel||Jianfeng Tan (Intel)||This presentation gives an introduction on virtio_user with vhost-kernel backend as a new exceptional path to exchange packets with kernel networking stack. This method provides benefits in: (1) maintenance, all kernel modules needed by this solution, vhost and vhost-net are upstreamed and extensively used kernel module; (2) features, vhost-net is born to be a networking solution, which has lots of networking related features, like multi queue, tso, multi-seg mbuf, etc; (3) performance, similar to KNI, kthread-based forwarding does not require DPDK polling thread trapping into kernel space to receive/transmit packets.|
FD.io (Fast Data) is the first truly open data plane project focusing on data IO performance, scalability, efficiency, and programmable flexibility for networking and storage. FD.io provides a modular, extensible user space IO services framework that supports rapid development of high-throughput low-latency resource-efficient IO services. The design of FD.io is hardware, kernel, and deployment (bare metal, VM, container) agnostic. FD.io has been integrated with OpenStack Neutron, OpenDaylight, and Calico to provide a drop in upgrade for all of your dataplane needs.A key component of Fido is the Vector Packet Processing (VPP) library contributed at the foundation of the project. The commercial-ready code targetable to run on standard x86, ARM, and Power servers is already running in products on the market today. VPP when connected to DPDK for network I/O has shown to perform two orders of magnitude faster than currently available open source options implementing switching or routing workloads, reaffirming one of the core principles of FD.io: a focus on performance.
Join us at the FD.io Mini summit to hear and learn from FD.io community experts who will be sharing information about the projects, use cases, capabilities, integration between FD.io and OpenStack/ODL/OPNFV/Other communities, tools and many more exciting topics. This is a great opportunity for the OPNFV attendees to share their thought leadership and innovations at one of the industry’s premier events.
|1:30 - 1:50||FD.io Overview||Balaji Ethirajulu (Ericsson)|
|1:55 - 2:25||Vectorizing the container||Ray Kinsella (Intel)||Vector Packet Technologies such as DPDK and FD.io/VPP revolutionized software packet processing initially for discrete appliances and then for NFV use cases. Container based VNF deployments and it's supporting NFV infrastructure is now the new frontier in packet processing and has number of strong advocates among both traditional Comms Service Providers and in the Cloud. This session will provide an overview of how DPDK and FD.io/VPP are rising to meet the challenges of the Container dataplane. The discussion will provide an overview of the challenges, recent new features and what is coming soon in this exciting new area for the software dataplane, in both DPDK and FD.io/VPP!|
|2:30 - 3:00||Implementing a Virtual Network Service based on VPP||Binbin Xu/Jian Gu (ZTE)||Speaker name: Binbin Xu/Jian Gu
Speaker email: firstname.lastname@example.org email@example.com
|3:00 - 3:15||BREAK|
|3:20 - 3:50||Enabling high performance flexible 5G User Plane and SFC data plane in VPP||Danny Zhou & Hongjun Ni (Intel)||5G network needs a more flexible architecture. This will allow a service- or slice-specific grouping of network functions to logical entities, and the mapping of logical to physical architecture. Service Function Chaining provides an end-to-end ordered set of service functions. The presentation will show a flexible and configurable 5G User Plane and SFC Data Plane that includes below key elements: 1) GTP-U implementation and performance evaluation based on VPP. 2) Maps different 5G User Planes into one physical architecture. 3) Maps different SFC Data Planes into one physical architecture.|
|3:55 - 4:25||DPDK in CSIT – Power of Fusion||Qun Wan/Ray Kinsella (Intel)||The FD.io community aim to develop high-performance and programmable dataplane technology. CSIT is FD.io’s automated functional and performance test sub-project gifted to open source community by CISCO to ensure bugs and performance regressions are identified as part of the development process. DPDK is a foundation technology and a key dependency for FD.io which directly used in 3 FD.io key sub-projects including TLDK, T-REX and VPP and OSV’s consumption. A high-performance and bug-free DPDK is therefore a key and prioritized requirement. This presentation covers 3 topics: 1 Overview of CSIT technology on functional and performance test. 2 Projects Intel DPDK team proposed to CSIT: DPDK TESTPMD performance with CSIT; DPDK Layer 3 Performance with CSIT; IPSec Performance; NSH_SFC with CSIT; TLDK with CSIT. 3 Finally we will cover future direction and next steps, like Cryptodev and vHost.|
|4:30 - 5:00||Accelerating FD.IO/VPP crypto workload with DPDK Cryptodev Framework||Fan Zhang (intel)||A key aspect of providing secure and reliable cloud services like SDN and NFV is to securely protect the control and data plane traffic. However, nowadays it is not easy to protect network flows due to the increased network throughput, the faster rate of application development, and more and more complex network applications. In the work presented here we demonstrate the integration of the DPDK Cryptodev Framework with FD.IO/VPP to make use of the Intel® Performance Library and Intel® QuickAssist Technology (QAT) to accelerate VPP’s IPsec crypto workload. With the features provided by the DPDK Cryptodev Framework combined with the enriched functionalities of VPP the user can quickly deploy an efficient and secure IPsec application to fit their needs and effortlessly migrate it between software and hardware lookaside, or even physical to virtual environments.|