CS412 Introduction to Computer Networking

1
CS412 Introduction to Computer Networking
Telecommunication

• Chapter 5 Network Layer

2
Topics

• Design Issues
• Routing Algorithms
• Congestion Control
• Internetworking

3
Position of network layer
4
Network Layer Design Issues

• Store-and-Forward Packet Switching
• Services Provided to the Transport Layer
• Implementation of Connectionless Service
• Implementation of Connection-Oriented Service
• Comparison of Virtual-Circuit and Datagram
  Subnets

5
Store-and-Forward Packet Switching

• The environment of the network layer protocols.

6
Figure 19.3 Network layer in an internetwork
7
Figure 19.4 Network layer at the source
8
Figure 19.5 Network layer at a router
9
Figure 19.6 Network layer at the destination
10
Services Provided to Transport Layer

• Designing goals
• Independent of subnet technology
• Transport layer shielded from number, type, and
  topology of subnets
• Uniform network address numbering
• Two Types of Services
• Connectionless
• Connection-oriented

11
Implementation of Connectionless Service

• Routing within a diagram subnet.

12
Implementation of Connection-Oriented Service

• Routing within a virtual-circuit subnet.

13
Comparison of Virtual-Circuit and Datagram Subnets
5-4
14
Routing Algorithms

• Network layer software
• Deciding which output line an incoming packet
  should be transmitted to
• Datagram made for each packet
• VC made for new VC setup

15
Routing Algorithms

• Desirable Properties
• Correctness
• Simplicity
• Robustness
• Stability
• Fairness
• Optimality

16
Routing Algorithms - Optimality

• What to optimize?
• Minimizing mean packet delay
• Maximizing total network throughput
• Problem
• The above two are in conflicts
• Compromise
• Minimizing number of hops a packet must take from
  source to destination

17
Classes of Routing Algorithm

• Two major classes
• Non-adaptive
• Adaptive

18
Two Major Classes of Routing Algorithm

• Adaptive
• Algorithms differ in
• where to get information, e.g.,
• Locally
• From adjacent routers
• From all routers
• when to change routes, e.g.,
• Every T sec
• When the load changes
• When the topology changes
• what metric used for optimization, e.g.,
• Distance
• Number of hops
• Estimated transit time

19
Routing Algorithms to be Studied

• Static (i.e., nonadaptive) routing
• Shortest path routing
• Flooding
• Dynamic (i.e., adaptive) routing
• Distance vector routing
• Link state routing

20
Optimality Principle

• If router J is on the optimal path from router I
  to router K, then the optimal path from J to K
  also falls along the same route.
• Proof?

K
J
I
21
Sink Tree

• Direct consequence of optimality principle
• Subnet graph
• Node router
• Arc link
• The set of optimal routes from all sources to a
  given destination form a tree rooted at the
  destination
• Might not be unique
• Goal of routing algorithms
• Discover and use the sink tree for all routers

22
Sink Tree

• Example
• Distance metric number of hops

Subnet
Sink tree
23
Shortest Path Routing

• Given a pair of routers, find the shortest path
  between them
• Subnet graph
• Node router
• Arc link
• Labels of arcs
• Function of factors
• Weighting function changed for different criteria

24
Dijkstra's Algorithm

• Each node labeled with
• (distance from source, best known path)
• Initially
• distance infinity
• best known path unknown
• Label might change to reflect better paths
• Node is either tentative or permanent
• Initially tentative
• As a path from source to that node discovered,
  label becomes permanent and never gets changed

25
(No Transcript)
26
Dijkstra's Algorithm

• How do we find the path?
• The algorithm given in the book works from
  destination to source
• Why?

27
Flooding

• Every incoming packet is sent out on every
  outgoing line except for the input line
• Problem
• Large number of packets are generated
• Solutions
• Hop counter
• Avoiding duplicates
• Selective flooding

28
Flooding - Conclusion

• Optimal
• Shortest path is always chosen
• ? No other algorithm can produce a shorter delay
• Robust
• Not practical in most applications
• Useful in some applications
• Military application robustness
• Distributed database applications
• Concurrent update
• Wireless networks
• Metric of other routing algorithms

29
Distance Vector Routing

• Table in each router
• Giving
• Best known distance to each destination
• Which line to use to get there
• Indexed by each router
• Each entry contains two parts
• Preferred outgoing line to use for that
  destination
• Estimate of "distance" to go there
• Best known distance to each destination
• Updated by exchanging information with neighbors

30
Distance Vector Routing

• Router
• Knows "distance" to each neighbor
• Sends list to each neighbor every T msec
• ? Receives lists from neighbors every T msec
• If neighbor X knows
• Distance from X to I is XI, and
• Distance from the router to X is m
• then delay from router to I via X is (XI m)
• Performs calculation for each neighbor to find
  the best
• Old table not used in calculation

31
(No Transcript)
32
Distance Vector Routing

• Count-to-Infinity Problem
• Good news spread fast, bad news leisurely
• Infinity has to be defined

33
Link State Routing

• A router
• Discovers its neighbors and learn their network
  addresses - HELLO
• Measures the delay or cost to each neighbor -
  ECHO
• Constructs a packet telling all it has just
  learned link state packet
• Sends this packet to all other routers flooding
  w/ duplicate avoidance
• Computes the shortest path to every other router
  Dijkstras algorithm

34
Link State Routing

• In effect
• Complete topology and all delays are
  experimentally measured and distributed to every
  router
• Dijkstra's algorithm is applied to find the
  shortest path

35
Link State Packets
36
Hierarchical Routing

• Problem
• Network size grows ? Routing tables grow
• Solution
• Hierarchical routing
• Idea
• Routers are divided into "regions"
• Router knows detail about routing within its
  region
• Router knows nothing about internal structure of
  other region

37
Path length from 1A to 5C?
38
Congestion Control

• Congestion
• Too many packets present in the subnet
• Effects
• Performance degraded
• Packet lost

39
Congestion and Network Performance
(Could be achieved by congestion control)
(Without congestion control)
40
Causes of Congestion

• Causes
• Too many packets need an output line
• ? queuing
• Problem not enough memory ? packets dropped
• Solution(?) adding more memory
• New problem timeout and retransmit ? worse
• Slow processors
• Low bandwidth lines
• Congestion tends to feed upon itself and become
  worse

41
Congestion Control v.s. Flow Control

• Flow control relates to traffic between two
  machines, while congestion control is more global
• Examples
• No congestion, flow control needed
• Flow control not needed, congestion
• Confusion
• Congestion control might send back "slow down"
  messages
• ? Not an acknowledgement

42
Congestion Control - General Principles

• Two groups
• Open loop
• Closed loop feedback loop
• Explicit feedback
• Implicit feedback
• Source deduces existence of congestion by making
  local observation
• E.g., time needed for acknowledgement to come back

43
Congestion Control - General Principles

• Closed Loop - Approach
• Monitor the system to detect when and where
  congestion occurs
• Pass this information to places where actions can
  be taken
• Adjust system operation to correct the problem

44
Congestion Control

• Congestion (Load gt Resources)
• Solutions
• Increase resources
• Decrease load

45
Congestion Prevention Policies

• Policies that affect congestion.

5-26
46
Congestion Control in VC Subnets

• Admission control
• Alternative routes
• Resource reservation
• Based on negotiated agreement

47
Congestion Control in VC Subnets
48
Choke Packets

• Used in both VC and datagram subnets
• Approach
• Each router monitors output line utilization
• Threshold for "warning state"
• A receiving router
• Checks packet to see if output line in warning
  state
• If yes then
• send a "choke packet" back to source host
• original packet tagged and forwarded

49
Choke Packet

• Source, upon receiving a choke packet
• Reduces traffic by a percentage after receiving
  choke packet
• Choke packet referred to same destination is
  ignored for a fixed time interval
• After time interval expired, listens
• If choke packet received then
• goto the step of reducing traffic
• else
• increase traffic

50
Choke Packet

• Typically
• First choke packet causes data rate reduced to
  50, then 25,
• Traffic is increased in smaller increments
• Why?

51
Hop-by-Hop Choke Packets

• Problem in high speed and long distance ? slow
  reaction
• Solution
• Hop-by-hop choke packets
• Buffers needed in routers
• Effects
• Quick relief at the price of more buffers

52
Hop-by-Hop Choke Packets

• (a) A choke packet that affects only the source.
• (b) A choke packet that affects each hop it
  passes through.

53
Load Shedding

• Discard whatever cannot be handled
• Which packets to drop?
• Application-dependent
• Priorities

54
Load Shedding

• Strategies
• Wine or milk
• Priority
• Priority classes
• Coupled with traffic shaping token bucket
• ? Packet without token sent with lowest priority
• Allowing VC set up with exceeding specification
• ? Contingent on low priority
• Header field needed (ATM 1bit)
• Discard as early as possible!

55
Jitter Control

• Jitter variation in packet arrival time
• Necessary for audio/video transmission
• Control
• Delay jitter is bounded by computing expected
  transit time for each hop along path
• Packet is checked for behind/ahead schedule
• Behind sent out ASAP
• Ahead held for a while
• Usually buffering is used to eliminate jitters

56
Internetworking

• A collection of interconnected networks.

57
How Networks Differ
58
How Networks Can Be Connected
59
Concatenated Virtual Circuits

• A sequence of VCs is set up from source through
  one or more gateways to destination
• Each gateway maintains tables of information of
  VCs
• Best when all networks have roughly the same
  properties

60
Concatenated Virtual Circuits
61
Connectionless Internetworking

• Packets are not required to follow same sequence
  of connections
• Routing per packet
• ? Possibly higher bandwidth than VC
• Packets might arrive out of order
• Problems
• Format
• Addressing

62
Connectionless Internetworking
63
Tunneling

• Encapsulating packets of a protocol in the
  payload of packets of another protocol
• Useful in
• Internetworking
• VPN
• IPv4 to IPv6 transition

64
Tunneling
65
Internetwork Routing

• AS (Autonomous System)
• Each network in internetwork independent of
  others
• Two-Level Routing
• Interior gateway protocol
• Exterior gateway protocol
• Differences between interior and exterior
  routings
• International boundaries
• Cost
• Quality of service

66
Fragmentation

• Why maximal packet size
• Hardware
• OS
• Protocols
• Standards
• Retransmission reduction
• Prevention of one packet occupying channel too
  long

67
Fragmentation

• Maximum packet size different in different
  protocol
• Examples ATM 48 bytes, IP 65,515 bytes
• Problem
• Large size packets want to go through network of
  smaller packet size
• Solution Fragmentation
• Allowing gateways to break packets into
  fragments, each as a separate internet packet
• Problem Easy to break but difficult to put back
  together

68
Fragmentation Strategies

• Transparent fragmentation
• Nontransparent fragmentation

Transparent fragmentation
G5
Nontransparent fragmentation
69
Transparent Fragmentation

• Properties
• Fragmentation transparent to subsequent networks
• Same exit gateway in each network
• Procedure
• Fragmentation
• Each fragment is addressed to same exit gateway
• Fragments are combined at exit gateway

70
Transparent Fragmentation

• Simple
• Problems
• Combination
• Performance
• Overhead on repeated assembling

71
Nontransparent Fragmentation

• Properties
• Combination done at destination host
• One or more exit gateway(s)
• Procedure
• Fragmentation
• Each fragment is addressed to exit gateway(s)
• Fragments are combined at destination host
• Higher performance in datagram model
• Problems
• Every host must be able to reassemble
• Overhead in packet header

72
Nontransparent Fragmentation

• Numbering
• Required for assembly
• Assembly approach
• Elementary fragment size defined
• Fragments of a packet
• maximum packet size of another network except
  for the last one
• Header fields
• Packet number
• Number of first elementary fragment
• End-of-packet bit

73
(No Transcript)