Routing und Flow Control im Internet der Zukunft. Routing and Flow Control in the Future Internet

Transcription

1 Institute of Computer Science Department of Distributed Systems Prof. Dr.-Ing. P. Tran-Gia Routing und Flow Control im Internet der Zukunft www3.informatik.uni-wuerzburg.de

2 Outline Two major problems of routing in the Internet Depletion of available IPv4 addresses Solution: IPv6 Interworking IPv6 IPv4 Deployment Growth of the routing tables in the DFZ Causes Solutions: principles of future Internet routing Flow control in the future Internet Pre-congestion notification (PCN) Admission control and flow termination Conclusion 2

3 Depletion of Free IPv4 Address Pool IANA (Internet Assigned Numbers Authority) Projected depletion of unallocated IPv4 address pool: IPv4 Address format: 4 bytes ~ addresses 8,4 addresses per km 2 earth surface Structure: IPv6 Address format: 16 bytes ~ addresses 6, addresses per mm 2 earth surface Structure: 2001:DB8:0:0:8:800:200C:417A Prefix notation: /16: 16 bits prefix (~ address block) Interworking problems IPv6 addresses unknown to legacy applications, hosts, and routers Dual-stack (IPv4 and IPv6) required 3

4 IPv4 IPv6 Interworking Principles: Tunneling IPv6 traffic tunneled through IPv4 networks IPv6 IPv4 IPv6 A X Y B B Data Y B Data B Data 4

5 IPv4 IPv6 Interworking Principles: Address Conversion Conversion between IPv4 and IPv6 addresses :0:0:0:0:ffff:Hex( ) Applicable only to actual IPv4 addresses Conversion between IPv4 headers and IPv6 headers Stateless IP/ICMP translation (SIIT) IPv4 IPv6 IPv4 5

6 IPv4 IPv6 Interworking Principles: NAPT Problem Real IPv6 address not convertible into IPv4 address Network address port translation (NAPT) IPv4 border router converts From IPv6 address and port Into other IPv4 address and port and back Example IPv6 NAPT IPv4 Request Response [A]:1234 [C]:80 IPv6 NAPT IPv4 B:5678 C:80 src dst [A]:1234 B:5678 src dst [C]:80 [A]:1234 C:80 B:5678 6

7 Planned and Actual Deployment of IPv6 Observation IPv6 hardly adopted Limited reachability for early adopters Other partial solution to IPv4 address depletion Private networks behind NATs 10/8, /12, /16 Source: presentation by G. Huston and G. Michalson (APNIC) at RIPE 56 in Berlin, May 2008 Planned deployment of IPv6 Actual deployment of IPv6 7

8 IPv4 Outage Experiment at IETF71 IPv4 outage experiment at IETF71 in Philadelphia ( ) IPv6 Internet is only a very small fraction of IPv4 Internet Most portals do not offer services over IPv6 8

9 The Internet: a Network of Networks local ISP Tier 3 ISP Tier-2 ISP local ISP local ISP Tier-2 ISP local ISP Tier 1 ISP NAP local ISP Tier 1 ISP Tier-2 ISP local ISP Tier 1 ISP Tier-2 ISP local ISP Tier-2 ISP local ISP 9

10 Basic BGP Information BGP information / / /22 AS-Path: AS338, AS20978 AS-Path: AS574, AS231, AS339, AS448 AS-Path: AS574, AS1079, AS2098, AS

11 Problem 2: Growth of Routing Table Sizes in the DFZ IPv4 FIB entries from (AS2) 11

12 Causes for Increasing FIB Sizes in DFZ (1) Provider independent addressing Longest prefix match Maximum length of propagated prefixes: / /16 DFZ / /23 Provider A Provider B /23 x 12

13 Causes for Increasing FIB Sizes in DFZ (2) Multihoming / /23 DFZ / /23 Provider A Provider B /23 13

14 Causes for Increasing FIB Sizes in DFZ (3) Traffic engineering / / /24 DFZ / / /24 Provider A Provider B /23 Incoming VoIP Incoming data 14

15 Causes for Increasing FIB Sizes in DFZ (4) Countermeasure against prefix hijacking Announcement of longer prefixes than necessary E.g. YouTube prefix hijacking incident by Pakistan Telecom ( ) Source: RIPE /22 AS /24 AS36561 AS17557 YouTube Pakistan Telecom 15

16 Solution 1: Tweaking the Current Interdomain Routing (1) Aggregation proxies Core router-integrated overlay (CRIO) X.Y.0/22 Statically configured tunnels X.Y.0/22 X.Y.0/22 Aggregation proxy announces short prefixes X.Y.0/22 The aggregation proxy announces a short prefix instead of many long prefixes. Packets addressed to the long prefixes are routable in the DFZ They are forwarded to the aggregation proxy which tunnels them to their destination network. X.Y.0/24 X.Y.1/24 X.Y.2/24 X.Y.3/24 16

17 Solution 1: Tweaking the Current Interdomain Routing (2) Retain long prefixes and provide lookup system for direct tunnels Tunneling route reduction protocol (TRRP) Lookup system for non-routable addresses X.Y.Z/24 Some long prefixes are not announced to BGP, therefore, they are not routable in the DFZ. The lookup system provides a router for them in the destination AS such that corresponding packets can be tunneled, decapsulated, and forwarded from there to their destination via intradomain routing. Border router with routable address X.Y.Z/24 17

18 Solution 2: Locator/Identifier Split Separation of IP addresses Identifier Locator Mapping function Identifier locator Objective Limit growth of routing tables Open issues Mapping system Exact implementation of Loc/ID Mapping service Provider X B A Data B Locator(B) Provider Y 18

19 Incremental Deployment of Loc/ID for the Internet Mapping service supported by local caches Locator ID separation protocol (LISP) Cisco s proposal within RRG of IRTF Gateways A 1 2 B Locators Identifiers 3 C 4 D Local routing domain Global routing domain Communication 1 4: 1 sends packet with address 4 to A, A sends packet with address D4 to D, D sends packet with address 4 to 4. 19

20 Interworking between the Legacy and the Future Internet Communication 1 B: 1 sends packet with address B to A, A sends packet with address B to B. Mapping service supported by local caches Global routing domain and legacy Internet Local routing domain 1 A Gateway C Proxy gateway B Legacy node Communication B 1: B sends packet with address 1 to C, C sends packet with address A1 to A, A sends packet with address 1 to 1. 20

21 Clean Slate Approach for Loc/ID Identifier (2) Local locator (LL(2)=b) Local mapping service b 2 a 1 Local mapping service LL(2)=b ID=2 Data 21

22 Clean Slate Approach for Loc/ID Identifier (2) Local locator (LL(2)=b) Local mapping service Global locator (GL(3)=C) Global mapping service A B C a f 1 b c d e 3 Local mapping service Global mapping service LL=b LL=d LL=c LL=e GL(3)=C ID=3 LL for next jump to C added using local routing tables LL(3)=f ID=3 Data LL(3)=f added by ingress node using local mapping service Data 22

23 Solutions for Improved Scalability Locator ID separation protocol LISP Different mapping implementations Distributed hash table LISP-DHT Alternative, logical topology LISP-ALT Content overlay network service LISP-CONS A not-so-novel EID to RLOC database LISP-NERD A practical tunneling architecture efit-apt Six/One Router with DNS-based resolution system Six/One Dynamic internetworking architecture DYNA Tunneling route reduction protocol TRRP Internet vastly improved plumbing Ivip Host identity protocol architecture HIP Global, site, and end-system address elements GSE Node identity interworking architecture Hierarchical routing architecture HRA New inter-domain routing architecture NIRA IP with virtual link extension IPvLX Core router-integrated overlay CRIO Geographically informed inter-domain routing GIRO On Compact Routing for the Internet 23

24 Pre-Congestion Notification (PCN) Flow Control for the Future Internet Simple support for quality of service (QoS) No per-flow states inside a network Admission control Proactive: keep traffic load low to avoid congestion High priority transport only for explicitly admitted flows Block further flows if traffic load is already high Flow termination Terminates some admitted flows Only for exceptional cases Reactive: reduce traffic load if it is too high due to an accicent 24

25 Pre-Congestion Notification (PCN) Concept PCN rate r(l) on link l Supportable rate SR(l) Admissible rate AR(l) 0 Pre-congestion type AR-precongestion SR-precongestion No precongestion Impact on AC and FT Block new flows Terminate some admitted flows Block new flows Admit new flows 25

26 Edge-to-Edge Pre-Congestion Notification (PCN) Source PCN Domain Destination End-to-end resource signalling End-to-end flow S PCN ingress node S S/MM MM S/MM PCN egress node RSVP S Capacity Overprovisioning S Router with signalling functionality MM Router with metering & marking functionality 26

27 End-to-End Pre-Congestion Notification (PCN) Source PCN Domain Destination End-to-end flow MM MM MM MM MM MM MM Router with metering & marking functionality 27

28 Conclusion Pre-congestion notification (PCN) Packet marking Admission control Flow termination Edge-to-edge and end-to-end PCN Two major problems in today s routing Depletion of available IPv4 address pool Growth of routing tables IPv6 Interworking methods with IPv4 No incentive for early adopters Hardly used Loc/ID split Promising design principle for routing scalability Incremental deployment e.g. LISP Clean slate Loc/ID What s routing like in the Internet in 2020? 28