Top Banner
Cisco Public 1 © 2014 Cisco and/or its affiliates. All rights reserved. IPv6 performance bonus? Eric VYNCKE, Cisco, [email protected] Marcel ENGUEHARD, École Polytechnique September 2014
9

Cisco Public 1 © 2014 Cisco and/or its affiliates. All rights reserved. Eric VYNCKE, Cisco, [email protected] Marcel ENGUEHARD, École Polytechnique September.

Dec 24, 2015

Download

Documents

Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: Cisco Public 1 © 2014 Cisco and/or its affiliates. All rights reserved. Eric VYNCKE, Cisco, evyncke@cisco.com Marcel ENGUEHARD, École Polytechnique September.

Cisco Public 1© 2014 Cisco and/or its affiliates. All rights reserved.

IPv6 performance bonus?Eric VYNCKE, Cisco, [email protected] ENGUEHARD, École Polytechnique

September 2014

Page 2: Cisco Public 1 © 2014 Cisco and/or its affiliates. All rights reserved. Eric VYNCKE, Cisco, evyncke@cisco.com Marcel ENGUEHARD, École Polytechnique September.

© 2014 Cisco and/or its affiliates. All rights reserved. Cisco Pubic 2

Web probe to web servers

HTTP HEAD request

TCPDUMP+

Java

Page 3: Cisco Public 1 © 2014 Cisco and/or its affiliates. All rights reserved. Eric VYNCKE, Cisco, evyncke@cisco.com Marcel ENGUEHARD, École Polytechnique September.

© 2014 Cisco and/or its affiliates. All rights reserved. Cisco Pubic 3

Web probe to web servers: locations

Page 4: Cisco Public 1 © 2014 Cisco and/or its affiliates. All rights reserved. Eric VYNCKE, Cisco, evyncke@cisco.com Marcel ENGUEHARD, École Polytechnique September.

© 2014 Cisco and/or its affiliates. All rights reserved. Cisco Pubic 4

Measurement results: global results

Distribution of log( ) for each source-destination pair from our web probe

Page 5: Cisco Public 1 © 2014 Cisco and/or its affiliates. All rights reserved. Eric VYNCKE, Cisco, evyncke@cisco.com Marcel ENGUEHARD, École Polytechnique September.

© 2014 Cisco and/or its affiliates. All rights reserved. Cisco Pubic 6

Measurement results: HE effect• HE: most important IPv6 backbone provider

• Routeview.org Ashburn’s router BGP table presence of AS6939:44,8% of the IPv6 routes

16,9% of the IPv4 routes

Hurricane Electric’s network map

Page 6: Cisco Public 1 © 2014 Cisco and/or its affiliates. All rights reserved. Eric VYNCKE, Cisco, evyncke@cisco.com Marcel ENGUEHARD, École Polytechnique September.

© 2014 Cisco and/or its affiliates. All rights reserved. Cisco Pubic 7

Measurement results: HE effect in Singapore

RTT distribution from Singapore to mail.de (in µs)Paths in IPv4 and IPv6 from Singapore to mail.de

Page 7: Cisco Public 1 © 2014 Cisco and/or its affiliates. All rights reserved. Eric VYNCKE, Cisco, evyncke@cisco.com Marcel ENGUEHARD, École Polytechnique September.

© 2014 Cisco and/or its affiliates. All rights reserved. Cisco Pubic 8

Measurement results: Non-congruent routing• Geographically incoherent routing

Path taken on AS6453 (TATA Comm.) by a TCP connection between our Rackspace server and e-foia.uspto.govDistribution of IPv4 and IPv6 RTTs (in µs) between our server in Virginia and

e-foia.uspto.gov

Page 8: Cisco Public 1 © 2014 Cisco and/or its affiliates. All rights reserved. Eric VYNCKE, Cisco, evyncke@cisco.com Marcel ENGUEHARD, École Polytechnique September.

© 2014 Cisco and/or its affiliates. All rights reserved. Cisco Pubic 9

Time to improve BGP routing and peering• BGP peering is not optimal, either in v4 or v6

Lots of Ases are happy enough to peer with only 2 or 3 tier 1 backbone providers

Peering agreements are specific to v6

• Specific ASes’ topology influence routing too muchParticularly in intercontinental connections

Page 9: Cisco Public 1 © 2014 Cisco and/or its affiliates. All rights reserved. Eric VYNCKE, Cisco, evyncke@cisco.com Marcel ENGUEHARD, École Polytechnique September.

Thank you.