DPDK Accelerated Load Balancer

Slides:

Advertisements

Similar presentations

Building Fast, Flexible Virtual Networks on Commodity Hardware Nick Feamster Georgia Tech Trellis: A Platform for Building Flexible, Fast Virtual Networks.

Advertisements

CAS3SH3 Midterm Review. The midterm 50 min, Friday, Feb 27 th Materials through CPU scheduling closed book, closed note Types of questions: True & False,

Supercharging PlanetLab : a high performance, Multi-Application, Overlay Network Platform Written by Jon Turner and 11 fellows. Presented by Benjamin Chervet.

IUT– Network Security Course 1 Network Security Firewalls.

Keith Wiles DPACC vNF Overview and Proposed methods Keith Wiles – v0.5.

VIA and Its Extension To TCP/IP Network Yingping Lu Based on Paper “Queue Pair IP, …” by Philip Buonadonna.

CSE 190: Internet E-Commerce Lecture 16: Performance.

1 Network Packet Generator Characterization presentation Supervisor: Mony Orbach Presenting: Eugeney Ryzhyk, Igor Brevdo.

Embedded Transport Acceleration Intel Xeon Processor as a Packet Processing Engine Abhishek Mitra Professor: Dr. Bhuyan.

FreeBSD Network Stack Performance Srinivas Krishnan University of North Carolina at Chapel Hill.

Copyright © 2002 Wensong Zhang. Page 1 Free Software Symposium 2002 Linux Virtual Server: Linux Server Clusters for Scalable Network Services Wensong Zhang.

SEDA: An Architecture for Well-Conditioned, Scalable Internet Services

Christopher Bednarz Justin Jones Prof. Xiang ECE 4986 Fall Department of Electrical and Computer Engineering University.

Hosting Virtual Networks on Commodity Hardware VINI Summer Camp.

LiNK: An Operating System Architecture for Network Processors Steve Muir, Jonathan Smith Princeton University, University of Pennsylvania

Submitted by: Shailendra Kumar Sharma 06EYTCS049.

MIDeA :A Multi-Parallel Instrusion Detection Architecture Author: Giorgos Vasiliadis, Michalis Polychronakis,Sotiris Ioannidis Publisher: CCS’11, October.

Fast Multi-Threading on Shared Memory Multi-Processors Joseph Cordina B.Sc. Computer Science and Physics Year IV.

Design and Implementation of a Multi-Channel Multi-Interface Network Chandrakanth Chereddi Pradeep Kyasanur Nitin H. Vaidya University of Illinois at Urbana-Champaign.

Computers Operating System Essentials. Operating Systems PROGRAM HARDWARE OPERATING SYSTEM.

VTurbo: Accelerating Virtual Machine I/O Processing Using Designated Turbo-Sliced Core Embedded Lab. Kim Sewoog Cong Xu, Sahan Gamage, Hui Lu, Ramana Kompella,

Memory and network stack tuning in Linux:

Full and Para Virtualization

6.894: Distributed Operating System Engineering Lecturers: Frans Kaashoek Robert Morris

Low Overhead Real-Time Computing General Purpose OS’s can be highly unpredictable Linux response times seen in the 100’s of milliseconds Work around this.

Overview on Web Caching COSC 513 Class Presentation Instructor: Prof. M. Anvari Student name: Wei Wei ID:

1 Chapter 2: Operating-System Structures Services Interface provided to users & programmers –System calls (programmer access) –User level access to system.

An open source user space fast path TCP/IP stack and more…

Network Virtualization Ben Pfaff Nicira Networks, Inc.

NFP: Enabling Network Function Parallelism in NFV

Introduction to Operating Systems Concepts

Shaopeng, Ho Architect of Chinac Group

Balazs Voneki CERN/EP/LHCb Online group

New Approach to OVS Datapath Performance

Barracuda Link Balancer

BESS: A Virtual Switch Tailored for NFV

GPUNFV: a GPU-Accelerated NFV System

Arrakis: The Operating System is the Control Plane

Kernel Design & Implementation

Module 12: I/O Systems I/O hardware Application I/O Interface

University of Maryland College Park

Affinity Depending on the application and client requirements of your Network Load Balancing cluster, you can be required to select an Affinity setting.

CS 6560: Operating Systems Design

SCTP Handoff for Cluster Servers

Lesson Objectives Aims Key Words

Andy Wang COP 5611 Advanced Operating Systems

The Multikernel: A New OS Architecture for Scalable Multicore Systems

Securing the Network Perimeter with ISA 2004

Logo here Module 8 Implementing and managing Azure networking 1.

Task Scheduling for Multicore CPUs and NUMA Systems

NFP: Enabling Network Function Parallelism in NFV

F-Stack - a full user-space network service with DMM

OSDI ‘14 Best Paper Award Adam Belay George Prekas Ana Klimovic

NFP: Enabling Network Function Parallelism in NFV

Operating System Concepts

13: I/O Systems I/O hardwared Application I/O Interface

CS703 - Advanced Operating Systems

Virtio/Vhost Status Quo and Near-term Plan

Integrating DPDK/SPDK with storage application

CT 1306 Communication Networks Management Lab

Accelerate Vhost with vDPA

All or Nothing The Challenge of Hardware Offload

Presented by Neha Agrawal

Prof. Leonardo Mostarda University of Camerino

Chapter 13: I/O Systems I/O Hardware Application I/O Interface

Supporting Time-Sensitive Applications on a Commodity OS

Chapter 13: I/O Systems I/O Hardware Application I/O Interface

Interrupts and Interrupt Handling

Module 12: I/O Systems I/O hardwared Application I/O Interface

A Closer Look at NFV Execution Models

Presentation transcript:

DPDK Accelerated Load Balancer DPVS: open source HIGH PERFORMANCE l4 Load balancer BASED ON DPDK. Lei CHEN @ IQIYI.com, 2018.

Agenda Load Balancer Challenges Performance Bottleneck How to Achieve High Performance Architecture and Modules for DPVS Implementation Details and Examples New Features & Future Plan DPDK Summit China 2018

Load Balancer Challenges High Performance Concurrency, Throughput, … Cost (less machines) Various Use Cases Load Balancing (FNAT, DR, TUN) IDC Internet access (SNAT, GRE) Security Related Easy to Evolution Development & Customization. Maintain & Bug Fixing & Deployment. DPDK Summit China 2018

Why Kernel Based LB Not That Fast ? Kernel is the bottleneck ? (LVS/2.6.32) Context Switching Resource Sharing & Locking IRQ Storm Complicated Features (Deep call-stack) Kernel by-pass Tech Good aspect Performance, Easy Develop, Adoption, .. Bad aspect Expensive, No Multi-purpose, No Utilities, Stable and security. Pic from google maglev. Pic from mTCP. DPDK Summit China 2018

How to Achieve High Performance ? Kernel by-pass Tech L2/L3 Forwarding Load Balancer HFT Case “Share-nothing” CPU localization No share == no lock Avoid Context Switch Queue/CPU/Task bonding Disable scheduling Polling vs. Interrupting Zero Copy Pre-allocated Cache NUMA Awareness Prefetch, Huge-page … … DPDK Summit China 2018

DPVS: high performance LB based on DPDK Master/Worker Model NIC Queue/CPU Bonding User-land Lite IP-Stack Lockless for Key Data Lockless Inter-core Msg All-in-One Functional DPDK Summit China 2018

DPVS: Function Modules DPDK Summit China 2018

NIC Queues & CPU Bonding DPDK Summit China 2018

IDC Friendly: Virtual Devices DPDK Summit China 2018

QUIC Support: UOA and ConHash Connection IP/port changing Use CID for Schedule Some App need Client IP UOA IP Option Private Protocol DPDK Summit China 2018

Something new & future Plan Available Recently Consistent Hash UOA: get real UDP client IP QUIC: support CIP based schedule Tunnels: IP-in-IP, GRE Traffic Control (Linux Like, Lite Ver.) Plan Big NIC (25G/40G) & higher performance IPv6 Support Session Sharing SNAT: Traffic Isolation Session Logging DPDK Summit China 2018

Thanks ! https://github.com/iqiyi/dpvs DPDK Summit China 2018