Thursday, June 15th
09:00-10:00 AM
A-201: Network Acceleration 2 - High-Performance Systems (Networking Track)
Paper Title: Accelerating Open Virtual Switch (OVS) Using P4 and IPDK

Paper Abstract: Open Virtual Switch (OVS) is an open-source programmable software switch that coordinates network traffic across multiple virtual machines hosting multiple applications. Many data centers use it to virtualize network functionality (NFV). However, since OVS is software, it often runs slowly and uses a lot of processor resources. A new approach to accelerating it uses a combination of recent hardware and software developments, namely an infrastructure or data processing unit (IPU or DPU), the P4 packet processing language, and the open-source Infrastructure Programmer Development Kit (IPDK). The powerful combination takes advantage of the complete acceleration capabilities of the IPUs as shown in examples. Developers can use P4 to build more complex programs and map them to the IPU hardware as well. The results are hardware-agnostic and work with any P4-capable devices used to offload Open Virtual Switch and other more advanced constructs. Data centers will find the approach can accelerate network performance today and through the next generation of network requirements and device capabilities.

Paper Author: Deb Chatterjee, Network Acceleration Team Lead, Intel

Author Bio: Deb Chatterjee is a Sr Director of Engineering leading IPU software development at Intel. His team has been working on the hardware offload of virtual switches, storage, and crypto for several years and is currently exploring offloading of Kubernetes networking and service mesh. He has presented his work at OVS and P4 conferences and received many divisional awards within Intel. He was a founder of the open source initiative for an Infrastructure Programmers' Development Kit or IPDK (www.ipdk.io). Before joining Intel, he was Director of Software at Ericsson, where he was responsible for the Software Development Kit (SDK) for Ericsson's multi-core Network Processor. He earned his MS in computer networking from University of California at Santa Cruz (UCSC).