Advanced Internetworking
14 important questions on Advanced Internetworking
How to make Routing scale? And solve address exhaustion?
- hierarchical IP addresses: network + host
- subnetting: subnets are not known to outside world
- CIDR: only prefixes are advertised
Observation: still too many networks
- routing tables do not scale
- route propagation protocols do not scale
ADDRESS SPACE EXHAUSTION
- subnetting: do not assign whole class B networks
- CIDR: aggregate small class C networks in useful units
- DHCP: temporal assignment of addresses
Observation: IPv4 address space is exhausted anyway
- there are more than 4 billion nodes
What used to be the Global Internet Structure (1990)?
- backbone, service providers, end-user sites
- hierarchy improves scalability
* local router only has to know local networks and some default router
* core routers have to know everything, but this can be summarized in a limited number of prefixers
How did the route propagation work?
- AS corresponds to an administrative domain
- examples: ISP, company, backbone
- central authority assigns each AS a unique 16 bit number (ASN)
- each AS is a routing domain
Two level route propagation hierarchy
- interior gateway protocol (IGP): each AS selects its own IGP
- exterior gateway protocol (EGP): internet wide standard
- Higher grades + faster learning
- Never study anything twice
- 100% sure, 100% understanding
What are the popular interior gateway protocols?
route information protocol (RIP)
- developed for XNS, distributed with BSD Unix
- original ARPANET protocol
- based on hop count (cost =1)
Enhanced Interior Gateway Routing Protocol (EIGRP)
- cisco proprietary protocol
- able to deal with CIDR, optimization based on multiple metrics
LINK STATE ROUTING
Open Shortest Path First (OSPF)
- recent IETF standard
- rich set of extensions, load balancing, authentication
Intermediate System to Intermediate System (IS-IS)
- developed by DEC / ISO standard
- used in networks of large providers
- easily adapted to support other addresses
What was the global internet structure in 2000
Large corporations can be connected to multiple backbones. Providers have “political” peering arrangements
Topology?
- it is no longer a hierarchy!
Stub AS: has a single connection to one other AS
- carries local traffic only
Multihomed AS: has connections to more than one AS
- refuses to carry transit traffic
Transit AS: has connections to more than one AS
- carries both transit and local traffic
What are the Exterior Gateaway Protocols?
Goal
- find loop-free path to destination
- concerned with reachability, not optimality
Challenges:
- topology: the Internet is no longer a hierarchy
- scale: backbone router must be able to forward any packet
- cost: how to calculate meaningful cost from heterogeneous metrics?
- trust: peering points are configured manually (can we trust provider X?)
- policies: policy conflicts may lead to convergence problems
Exterior Gateway Protocol (EGP)
- distance-vector: neighbor routers periodically exchange their routing tables
- was designed for original tree-structured Internet
Border Gateway Protocol (BGP)
-current core Internet routing protocol (BGP-4)
What is the Border Gateway Protocol?
Each AS has one speaker and one or more border routers
- packets leave/enter AS via border routers
- BGP speaker advertises to other speakers: local networks, other reachable networks (transit AS only), path information
- speaker can cancel previously advertised paths
BGP-4 is a Path Vector protocol
- advertises complete paths: to enable policy decisions, to enable detection of loops
- supports subnets and CIDR: advertises classless prefix/length addresses
- very complex protocol involving multiple processes: receiving messages, processing routes, selecting best routes, applying policies, …
- running on top of explicit, reliable TCP connections (using port 179)
What are OSPF Routing Areas?
Link-state protocols further partition a domain in areas
- link-state advertisements stay in one area
Backbone Area 0
- traffic between areas always via Area 0
Area Border Router (ABR)
- R1, R2, R3
- also receive link-state advertisements
- consider all networks in area as if directly connected
- advertise summary information
Trade-off: scalability versus optimality
How does the Global Internet structure look like now
- Amazon, Google, Facebook, ...
- e.g. Google offering Web Search, Gmail, Youtube, App Engine, ...
Big players have their own private backbone networks
- to minimize extent to which client traffic traverses public Internet
Big players have settlement-free peering with multiple ISPs
- they don’t pay ISPs, but refuse to carry transit ISP traffic
- reverse-blocking possible? (ISP has to pay big player)
How does Google work?
Google has its own private backbone: B4
- WAN interconnecting server clusters grouped in data centers
- DWDM on dark fiber
- each cluster is logical AS
- BGP / IS-IS routing
- SDN based TE
- ECMP load balancing
Second network connects data centers to peering edge: B2
- using decentralized routing protocols (BGP, ...)
- Google has peering arrangements with (almost) all major ISPs
What are the IP version 6 features?
Features
- hierarchical 128-bit addresses (classless)
- multicast and anycast (see §4.2) *
- real-time services (see §6.5) *
- authentication and security (see chapter 8) *
- autoconfiguration *
- end-to-end fragmentation
- mobility support *
- multihoming
- protocol extensions
Simplified header
- 40-byte “base” header
- extension headers
Explain the IPv6 Packet Format
40-byte “base” header
- Version = 6
- TrafficClass and FlowLabel
- PayloadLen
- HopLimit
Extension headers
- NextHeader determines first extension header
- extension headers replace multiple fields of IPv4 header: Options, Protocol, fragmentation (Ident, Flags, Offset)
No checksum?
How transition from IPv4 to IPv6?
Overnight transition is not possible
- IPv4 and IPv6 will coexist for some time
Existing IPv4 hosts …
- can be upgraded to IPv6 (even if other hosts / routers not upgraded)
- can continue to use IPv4 (even if IPv6 installed on other nodes)
Required upgrades to handle IPv6
- DNS servers (see §9.3), routers & routing protocols, DHCP, …
Multiple transition strategies are possible ...
What are some transition strategies?
1. Dual-stack: two separate IPv4 and IPv6 stacks in devices
- select stack depending on IP version to use
- coexistence of IPv4 and IPv6 mechanisms: e.g. DHCP versus DHCPv6/autoconfig
2. Tunneling: encapsulate IPv6 packets inside IPv4 packets
- Configured tunneling: tunnel endpoint IPv4 address explicitly configured
- Intra-Site Automatic Tunnel Addressing Protocol (ISATAP): ISATAP tunneling interface: fe80::5efe:10.11.12.13
3. Header translation
- new IPv6-only devices communicate via NAT to IPv4-only devices
- extended NAT device translates header and addresses
The question on the page originate from the summary of the following study material:
- A unique study and practice tool
- Never study anything twice again
- Get the grades you hope for
- 100% sure, 100% understanding