Title: Multihoming Issues
 1Multihoming Issues on IPv6 networks
APRICOT2002 Dongkyun Kim, Ilsun Whang E-mail  
mirr_at_kreonet2.net 
 2Contents
-  Introduction to KREONet2 6Bone 
-  About IPv6 
-  Interdomain Routing Issues 
-  IPv6 Multihoming Requirements 
-  Related Mechanisms 
-  IPv6 Multihoming issues 
-  Summary 
3Introduction to KREONet2 6Bone 
-  KREONet2 
-  stands for Korea Research Environment Open 
 Network 2
-  National RD network, run by KISTI, supported by 
 Korean government.
-  Upgraded name of KREONet and next generation 
 research network.
-  KREONet2 6Bone 
-  KREONet2 sTLA  2001320/35 
-  Peering with 6TAP, KOREN, and 6NGIX. 
-  Web, DNS, and Tunnel Broker service is ongoing. 
-  Plan to provide end-to-end IPv6 service this 
 year.
4Introduction to KREONet2 6Bone 
KIX
STAR TAP 6TAP
200Mbps
10Mbps
Chunchon
Seoul
Seoul
45Mbps
Inchon
45Mbps
Suwon
155Mbps
IMNet
KREONet2 6Bone (AS17579)
256Kbps
Chonan
Chungju
10Mbps
45Mbps
1Gbps
Daejon
Daejeon
16Mbps
Internet
45Mbps
45Mbps
4Mbps
Pohang
Daegu
Jeonju
Ulsan
2Mbps
45Mbps
45Mbps
45Mbps
Kwangju
Pusan
Changwon
4 Mbps
Jeju 
 5About IPv6 
-  Background 
-  Rising crisis of IPv4 address depletion 
-  Non-hierarchical routing system problem 
-  Growth of the applications 
-  Temporary solutions 
-  - Renumbering, CIDR, NAT, DHCP 
-  Benefits 
-  Increased IP address size 
-  Increased addressing hierarchy support 
-  Simplified host adddressing 
-  Simpler Autoconfiguration of address 
-  And so on
6About IPv6 
-  The need for further development 
-  The multihoming problem 
-  Less-rigid structure for global unicast 
 addresses
-  Anycast details 
-  And so on
7Interdomain Routing Issues
-  Growth of interdomain routing tables
By Geoff Huston (2001.1) 
 8Interdomain Routing Issues
-  More specific prefixes advertisement
KREONet2 NOC, http//measurement.ipv6.re.kr/bgp/ 
 9Interdomain Routing Issues
KREONet2 NOC, http//measurement.ipv6.re.kr/bgp/ 
 10Interdomain Routing Issues
By Geoff Huston, http//www.telstra.net/ops/bgp/ 
 11IPv6 Multihoming Requirements
-  Scalability 
-  Redundancy 
-  Physical / logical link failure 
-  Routing protocol failure 
-  Transit provider failure 
-  Load Sharing 
-  Inbound traffic load sharing 
-  outgoing traffic load sharing 
-  Performance 
-  e.g. Avoiding long-term congestion between 
 providers
12IPv6 Multihoming Requirements
-  Policy 
-  e.g. cost, acceptable use conditions, etc.) 
-  Provider selection for a certain traffic class 
 or application
-  Simple operations and management 
-  Current multihoming solution is quite 
 straightforward to deploy and maintain.
-  Transport-layer Survivability 
-  providing re-homing transparency 
13IPv6 Multihoming Requirements
-  Security Considerations 
-  Denial Service attack and spoofing attacks are 
 possible, so multihomed sites must be protected
 against such attacks at least as well as
 single-homed sites
-  More considerations should be taken into 
 possible tunnel link
-  Encryptions 
-  Packets may travel an unwanted path, otherwise 
 secondary links are configured with care
14Related Mechanisms
-  IPv6 multihoming with route aggregation 
-  draft-ietf-ipngwg-ipv6multihome-with-aggr-01 
 (now obsolete)
-  a site multi-homed to more than one ISPs 
-  Provider level multihoming technique 
-  IPv6 multihoming support at site exit routers 
-  RFC 3178 
-  a site multi-homed to more than one ISPs using 
 different site border routers
-  Site level multihoming technique
15Related Mechanisms
-  Multihoming Mechanism with route aggr.
Inbound Traffic 1. Customer-A  Addr-1-A, 
specific route advertising (ISP-1, ISP-2) 2. 
ISP-2 advertises Addr-1-A to ISP-1 only. 3. 
ISP-1 advertises Addr-1-A aggregation block to 
ISP-3, ISP-4
ISP-4
ISP-3
3
3
ISP-2
ISP-1
2
Outgoing Traffic - Default route advertising  
ISP-1 and ISP-2 - Or selective set of specific 
prefixes advertising According to the requirement 
of customer-A
1
1
Customer-A 
 16Related Mechanisms
-  Multihoming Mechanism with route aggr.
Load Sharing 1. Inbound Traffic - Achieved by 
careful controlling route policy of ISP-1 - 
Using IGP Metric, BGP route selection 2. Outbound 
Traffic - Achieved by controlling 
advertisement from ISP-1 and ISP-2
ISP-4
ISP-3
ISP-2
ISP-1
Redundancy 1. Link-1 failure - In  ISP-1 
-gt ISP-2 -gt Customer-A - Out  Customer-A -gt 
ISP-2 2. Link-2 failure - In  ISP-1 -gt 
Customer-A - Out  Customer-A -gt ISP-1
Link-1
Link-2
Customer-A 
 17Related Mechanisms
-  Most Suitable Environments for deploying 
-  In case that ISPs supporting multi-homed site 
 have direct connection to each other gt simple
 arrangement and improved aggregation
-  In case of needing good redundancy, not absolute 
 redundancy
-  Sites with limited to resource for onsite 
 sophisticated network administrators  Primary
 ISP takes the most part.
-  Sites that able to choose a robust ISP as 
 primary provider
18Related Mechanisms
-  IPv6 multihoming support at site exit routers 
-  Goals 
-  Scalability 
-  Redundancy 
-  Non-goals 
-  Choose the best exit link as possible 
-  Load balancing between multiple exit link
19Related Mechanisms
-  multihoming mechanism at site exit routers 
-  Establish secondary link using IP-over-IP tunnel 
-  Secondary link should be established through 
 different medium with primary link.
ISP-B
ISP-A
ISP-BR-A
ISP-BR-B
Secondary link
Multi-homed Site
E-BR-A, Pref-A
E-BR-B, Pref-B 
 20Related Mechanisms
-  multihoming mechanism at site exit routers 
-  Redundancy and Scalability 
Inbound Traffic 1. Strong Pref-A 2. Weak 
Pref-B 3. Strong Pref-B 4. Weak Pref-A
Outbound Traffic 1. Strong Default 4. Weak 
Default 3. Strong Default 2. Weak Default
1
2
3
4 
 21IPv6 Multihoming Issues
-  Scalability 
-  Routing table size has been a major issue (IPv4, 
 IPv6)
-  In IPv6, routing table size issue is more 
 serious negative effects on router memory usage,
 routing table lookup performance
-  Global routing table could be reduced using 
 aggregation hierarchy in IPv6, but the
 alternatives for multihoming are still be on the
 construction.
-  Redundancy 
-  inbound and outgoing link redundancy 
22IPv6 Multihoming Issues
-  Load Sharing 
-  inbound traffic load sharing 
-  Router Renumbering 
-  Source address selection between different 
 prefixes
-  Reliable link operation using tunneling 
-  Managing and operating tunneled link 
-  In case of long-term primary link down, usage of 
 other ISPs primary link is more efficient.
-  Marking Pref-A prefix as deprecated  no new 
 connection allowed
-  New connection requests are to be done using 
 newly assigned Pref-B
-  RRP makes this work automatic, but be very 
 cautious of the links flap
23Summary
-  Global routing table issues - Scalability 
-  Entries advertising more specific prefixes of 
 aggregates
-  Prefix distribution 
-  Growth of AS numbers 
-  IPv6 Multihoming Issues 
-  Need mechanism that can support scalability, 
 load sharing, redundancy.
-  Need to achieve better reachability without 
 impacting worldwide routing table size issues.
24Thank You !
- Contact Info 
- Director 
- Il-sun Whang 
- - his_at_kreonet2.net 
- Engineers 
- Dongkyun Kim 
- - mirr_at_kreonet2.net 
- Hyeakro Lee 
- - leehr_at_kreonet2.net