1 Building a Fast, Virtualized Data Plane with Programmable Hardware Bilal Anwer Nick Feamster.

Slides:



Advertisements
Similar presentations
You have been given a mission and a code. Use the code to complete the mission and you will save the world from obliteration…
Advertisements

EE384y: Packet Switch Architectures
Advanced Piloting Cruise Plot.
Copyright © 2003 Pearson Education, Inc. Slide 1 Computer Systems Organization & Architecture Chapters 8-12 John D. Carpinelli.
Chapter 1 The Study of Body Function Image PowerPoint
1 Copyright © 2013 Elsevier Inc. All rights reserved. Chapter 4 Computing Platforms.
Processes and Operating Systems
1 Copyright © 2013 Elsevier Inc. All rights reserved. Appendix 01.
Properties Use, share, or modify this drill on mathematic properties. There is too much material for a single class, so you’ll have to select for your.
Building Fast, Flexible Virtual Networks on Commodity Hardware Nick Feamster Georgia Tech Trellis: A Platform for Building Flexible, Fast Virtual Networks.
Network Virtualization Nick Feamster, Georgia Tech Lixin Gao, UMass Amherst Jennifer Rexford, Princeton NSF NeTS-FIND PI Meeting.
1 Resonance: Dynamic Access Control in Enterprise Networks Ankur Nayak, Alex Reimers, Nick Feamster, Russ Clark School of Computer Science Georgia Institute.
Multihoming and Multi-path Routing
Business Transaction Management Software for Application Coordination 1 Business Processes and Coordination.
1 Introducing the Specifications of the Metro Ethernet Forum MEF 19 Abstract Test Suite for UNI Type 1 February 2008.
Jeopardy Q 1 Q 6 Q 11 Q 16 Q 21 Q 2 Q 7 Q 12 Q 17 Q 22 Q 3 Q 8 Q 13
Jeopardy Q 1 Q 6 Q 11 Q 16 Q 21 Q 2 Q 7 Q 12 Q 17 Q 22 Q 3 Q 8 Q 13
Title Subtitle.
Properties of Real Numbers CommutativeAssociativeDistributive Identity + × Inverse + ×
Local Customization Chapter 2. Local Customization 2-2 Objectives Customization Considerations Types of Data Elements Location for Locally Defined Data.
My Alphabet Book abcdefghijklm nopqrstuvwxyz.
DIVIDING INTEGERS 1. IF THE SIGNS ARE THE SAME THE ANSWER IS POSITIVE 2. IF THE SIGNS ARE DIFFERENT THE ANSWER IS NEGATIVE.
FACTORING ax2 + bx + c Think “unfoil” Work down, Show all steps.
Addition Facts
Year 6 mental test 5 second questions
Year 6 mental test 10 second questions
Year 6 mental test 15 second questions Calculation Addition.
Around the World AdditionSubtraction MultiplicationDivision AdditionSubtraction MultiplicationDivision.
Break Time Remaining 10:00.
Copyright © 2009 EMC Corporation. Do not Copy - All Rights Reserved.
Debugging operating systems with time-traveling virtual machines Sam King George Dunlap Peter Chen CoVirt Project, University of Michigan.
Chapter 1: Introduction to Scaling Networks
PP Test Review Sections 6-1 to 6-6
Seungmi Choi PlanetLab - Overview, History, and Future Directions - Using PlanetLab for Network Research: Myths, Realities, and Best Practices.
ABC Technology Project
© 2006 Cisco Systems, Inc. All rights reserved. MPLS v MPLS VPN Technology Introducing MPLS VPN Architecture.
IP Multicast Information management 2 Groep T Leuven – Information department 2/14 Agenda •Why IP Multicast ? •Multicast fundamentals •Intradomain.
VOORBLAD.
Making Time-stepped Applications Tick in the Cloud Tao Zou, Guozhang Wang, Marcos Vaz Salles*, David Bindel, Alan Demers, Johannes Gehrke, Walker White.
Virtual Switching Without a Hypervisor for a More Secure Cloud Xin Jin Princeton University Joint work with Eric Keller(UPenn) and Jennifer Rexford(Princeton)
Copyright © 2012, Elsevier Inc. All rights Reserved. 1 Chapter 7 Modeling Structure with Blocks.
1 RA III - Regional Training Seminar on CLIMAT&CLIMAT TEMP Reporting Buenos Aires, Argentina, 25 – 27 October 2006 Status of observing programmes in RA.
Factor P 16 8(8-5ab) 4(d² + 4) 3rs(2r – s) 15cd(1 + 2cd) 8(4a² + 3b²)
Squares and Square Root WALK. Solve each problem REVIEW:
1..
Defect Tolerance for Yield Enhancement of FPGA Interconnect Using Fine-grain and Coarse-grain Redundancy Anthony J. YuGuy G.F. Lemieux September 15, 2005.
© 2012 National Heart Foundation of Australia. Slide 2.
Adding Up In Chunks.
Understanding Generalist Practice, 5e, Kirst-Ashman/Hull
Chapter 5 Test Review Sections 5-1 through 5-4.
SIMOCODE-DP Software.
GG Consulting, LLC I-SUITE. Source: TEA SHARS Frequently asked questions 2.
Addition 1’s to 20.
25 seconds left…...
Week 1.
We will resume in: 25 Minutes.
©Brooks/Cole, 2001 Chapter 12 Derived Types-- Enumerated, Structure and Union.
Essential Cell Biology
Intracellular Compartments and Transport
PSSA Preparation.
Essential Cell Biology
McGraw-Hill©The McGraw-Hill Companies, Inc., 2001 Chapter 16 Integrated Services Digital Network (ISDN)
NetSlices: Scalable Multi-Core Packet Processing in User-Space Tudor Marian, Ki Suh Lee, Hakim Weatherspoon Cornell University Presented by Ki Suh Lee.
Performance Evaluation of Open Virtual Routers M.Siraj Rathore
1 In VINI Veritas: Realistic and Controlled Network Experimentation Jennifer Rexford with Andy Bavier, Nick Feamster, Mark Huang, and Larry Peterson
Networking Virtualization Using FPGAs Russell Tessier, Deepak Unnikrishnan, Dong Yin, and Lixin Gao Reconfigurable Computing Group Department of Electrical.
Hosting Virtual Networks on Commodity Hardware VINI Summer Camp.
Presentation transcript:

1 Building a Fast, Virtualized Data Plane with Programmable Hardware Bilal Anwer Nick Feamster

2 Network Virtualization Network virtualization enables many virtual networks to share the same physical network resources. Many possible applications: –Hosting of multiple service provider networks –Experimentation –Running new protocols side-by-side with old ones

3 Fixed Network Infrastructure

4 Shared Infrastructure Networks have illusion of dedicated hardware.

5 Network Virtualization: Requirements Scalability –Support large number of networks (implies sharing) Performance –Support real traffic at line rate Flexibility –Support custom network services Isolation –Protection of networks from each other

6 Goal: Fast, Virtualized Data Plane Strawman approach: Software –Provides flexibility –…but poor performance and often inadequate isolation Our approach –Control plane in software –Data plane in hardware –Share hardware elements among virtual networks where possible

7 Virtualized Data Plane Router-1 Router-6 Router-2 Router-5 Router-3 Router-7 Router-8 SourceSink 2 Ethernet links 16 Ethernet links Router-4 Virtual router Router-4

8 Hardware-Based Virtualization Forwarding in hardware –faster than software –provides better isolation Sharing physical substrate amortizes cost –Unused hardware resources are already paid for Key challenge: Design must take advantage of both hardware and software –Requires interface between hardware and software –Requires identifying elements that can be shared among many virtual networks

9 Design Overview Control plane –two contexts –virtual environments in OpenVZ Interface to NetFPGA based on NetFPGA reference router

10 Talk Outline Implementation –Virtualization at Layer 2 –Fast forwarding –Resource guarantees per virtual network Preliminary Results –Performance & Efficiency Conclusion and Future Work

11 Virtualization at Layer 2 VRouter-1 VRouter-6 VRouter-2 VRouter-5 VRouter-3 VRouter-7 VRouter-8 Source VRouter-4 VMAC- VE Table Sink 00:11:22:33:44:55 0x1 00:11:22:33:44:55l

12 Layer-2 Virtualization: VMAC-VE Table VMAC-VE Table –provides virtualization at Layer 2 –maintains states for virtual Ethernet interfaces of each virtual environment Current implementation –Max. of four Ethernet interfaces per virtual router (currently limited by on-chip memory) –Max. of eight virtual routers working in parallel Hence, 32 Table Entries

13 Mapping the Virtual Forwarding Tables VMAC in packet determines the virtual network (and, hence, which CAMs to use)

14 Resource Guarantees CPU Isolation –Provided by using PCI-based NetFPGA card Bandwidth Isolation –Virtual networks are not affected by each other if they abide by their allocated bandwidth –What if user steps beyond allocated limited? Currently, no enforcement (limitation) Limit could be enforced at either ingress or egress

15 Evaluation What forwarding rates does the architecture achieve? How do these rates compare to the forwarding rate of the base hardware? How will the architecture scale with future hardware trends?

16 Experimental Setup

17 Forwarding Performance: Rates Forwarding Rate (000 pps) Packet Size (bytes) Packet forwarding rates are at least as good as Linux kernel. (~2.5x for small packets)

18 Forwarding Performance: Overhead Forwarding Rate (000 pps) Packet Size (bytes) Performance of up to eight virtual routers is equivalent to base router.

19 Efficiency Base router: 45% of logic, 53% of BRAM, 8.6M gates 8 Virtual Routers: 69% of logic, 87% of BRAM, 14.1M gates Virtual Routers Cards will support more virtual routers as Xilinx technology improves.

20 Future Work Adding support for forwarding tables on SRAM. Providing bandwidth isolation when users exceed allocated bandwidth. Providing an interface to each user for performance statistics, etc.

21 Summary: Fast, Virtualized Data Plane Scalable –Design is scalable (Off-chip FIB will allow more virtual data planes.) Fast –Current implementation has the same performance as base hardware Flexible –Support for custom control and data planes Provides Isolation –Virtual networks dont interfere with each other if traffic within limits

22 Conclusion Resource sharing in routers using programmable hardware is possible Hardware resource sharing provides improved isolation and packet forwarding rates than software based solution Current implementation achieves isolation and forwarding performance of native hardware without any overhead

23

24 Extra

25 Extra

26 Extra

27 Performance Overhead Tested with 1,2,3,4,5,6,7,8 virtualized data-planes working in parallel and for 64-byte sized packets The forwarding rate was same for all eight virtualized data configuration All eight configuration showed forwarding rate equal to base router forwarding rate for 64-byte sized packets