COMP 631: COMPUTER NETWORKS. Internet Routing. Jasleen Kaur. Fall Forwarding vs. Routing: Local vs. Distributed

1  

COMP 631: COMPUTER NETWORKS

Internet Routing

Jasleen Kaur

Fall 2014

2  

q  Both datagram and virtual-circuit based networks need to

know how to construct forwarding tables q  Forwarding:

Ø  Select an output port based on destination address and routing table

Ø  Simple and well-defined process performed locally at a given node

Ø  Forwarding table used while forwarding packets

§ eg: (network number, interface id, MAC address)

q  Routing:

Ø  Building the forwarding/routing table

Ø  A complex, distributed algorithm problem

Ø  Routing table may be a precursor to a forwarding table

§ eg: (network number, nextHopIP, CostToDestination)

Forwarding vs. Routing: Local vs. Distributed

3  

q 

Intra-domain routing ~ 100 routers

q 

Given:

Ø  Graph: where nodes are routers and edges are links

Ø  Cost: associated with each link

q 

Find:

Ø  Lowest-cost path between any two nodes

q 

Requirements:

Ø  Self-healing, traffic-sensitive, scalable

Intra-domain Routing: Formulation

Need dynamic, distributed algorithms!

Two classes: based on “distance-vector” and “link-state” 4 3 6 2 1 9 1 1 D A F E B C 4   q 

Each node:

Ø  Constructs a vector of distances to all other nodes §  Distance vector

Ø  Distributes to immediate neighbors

q 

Neighbors use the distributed information to update

their own distance vectors

q 

This distributed exchange-update-exchange should lead

to globally consistent distance vectors (and routing

tables)

5  

Distance Table Data Structure

q 

Each node has its own table with a...

Ø Row for each possible destination

Ø Column for each directly-attached

adjacent node (neighbor)

q 

Each table entry gives cost to reach

destination via that adjacent node

Ø Distance = Cost DE₍₎     A     B     C     D   A     1     7     6     4   B     14     8     9     11   D     5     5     4     2   Cost to Destination via" D e st in atio n "

DX_{(Y,Z)  =    distance  from  X  to  Y  via  Z  as  first  hop}  

=    c(X,Z)  +  minw{DZ(Y,w)}

 

w  =  {neighbors  of  Z}   6   DE₍₎     A     B     C     D   A     1     7     6     4   B     14     8     9     11   D     5     5     4     2   Cost to Destination via" D e st in a tio n"

Distance Table Example

DE_{(C,D)    =    c(E,D)  +  min} w{DD(C,w)}   =    2  +  2    =    4

 

A  loop?!   7 8 1 2 2 1 A E B C D

DE_{(A,D)    =    c(E,D)  +  min}

w{DD(A,w)}  

=    2  +  3    =    5

 

Loop!  

DE_{(A,B)    =    c(E,B)  +  min}

w{DB(A,w)}  

7  

Distance Table Example

7 8 1 2 2 1 DE₍₎     A     B     C     D   A     1     7     6     4   B     14     8     9     11   D     5     5     4     2   D e st in atio n " A E B C D DE₍₎     A     B     C     D   A,  1     D,  5     D,  4     D,  2   Adjacent Node to Use, Cost" D e stin a tio n "

q  The distance table

gives the routing table

Ø  Just take the minimum cost per

destination

Distance Table" Routing Table"

8  

Distance Vector Routing: Algorithm

q 

Iterative, asynchronous – each

local iteration caused by:

Ø  Local link cost change, or

Ø  Message from adjacent node that its least cost path to some destination has changed

q 

Distributed:

Ø  Each node notifies adjacent nodes only when its least cost path to some destination changes

Ø  Adjacent nodes then notify their adjacent nodes if this update changes a least cost path

wait for change in local

link cost or message from

adjacent node"

Each node:!

recompute distance table"

ifleast cost path to any

destination has changed,

9   DX      _{Y        Z}   Cost  via   D es t.   _Y   Z   2          83          7    

Distance Vector Algorithm: Example

DX_{(Y,Z)  =  c(X,Z)  +  min} w{DZ(Y,w)}   =  7  +  1  =  8

 

2! 1! 7! Z Y X DZ      _{X        Y}   Cost  via   X          7          ∞   Y          ∞        1   D es t.   DY      _{X        Z}   Cost  via   X          2          ∞   Z          ∞        1   D es t.   DX      _{Y        Z}   Cost  via   Y          2          ∞   Z          ∞        7   D es t.  

DX_{(Z,Y)  =  c(X,Y)  +  min}

w{DY(Z,w)}   =  2  +  1  =  3

 

Time   10   DY      _{X        Z}   Cost  via   D es t.   DX      _{Y        Z}   Cost  via   D es t.   DZ      _{X        Y}   Cost  via   D es t.  

Distance Vector Algorithm: Example

2! 1! 7! Z Y X DZ      _{X        Y}   Cost  via   X   Y   D es t.   X          7          3   Y          9          1   DY      _{X        Z}   Cost  via   X   Z   D es t.   X          2          8   Z          9          1   DX      _{Y        Z}   Cost  via   Y   Z   D es t.   Y          2          8   Z          3          7   DZ      _{X        Y}   Cost  via   X          7          ∞   Y          ∞        1   D es t.   DY      _{X        Z}   Cost  via   X          2          ∞   Z          ∞        1   D es t.   DX      _{Y        Z}   Cost  via   Y          2          ∞   Z          ∞        7   D es t.   Time  

11   via   DY      _{X        Z}   X   via   DZ      _{X        Y}   X   4 1 50 Z Y X

Distance Vector: Link Cost Changes

  When  a  node  detects  a  local  link  cost  change:    

The  nodes  updates  its  distance  table    

If  the  least  cost  path  changes,  the  node  no^fies  its   neighbors   “Good news travels fast” 1   via   DY      _{X        Z}   X          4          6   via   DZ      _{X        Y}   X        50        5   via   DY      _{X        Z}   X          1          6   via   DY      _{X        Z}   X          1          6   DZ      _{X        Y}  via   X        50        2   DZ      _{X        Y}   X        50        5   via   c(X,Y)"

changes" t0   t1   t2   terminates"Algorithm" t3   50        2  

1          3  

How long does convergence take?

12   via   DZ      _{X        Y}   X        50        7   DZ      _{X        Y}  via   X        50        9  

Distance Vector: Link Cost Changes

  Good  news  travels  fast,  but…    

  “Bad  news”  travels  slow!  

The  “count  to  infinity”  problem  

Routing Loop!! Does it Terminate?! ₄ 1 50 Z Y X 60   via   DY      _{X        Z}   X          4          6   via   DZ      _{X        Y}   X        50        5   via   DY      _{X        Z}   X        60        6   via   DY      _{X        Z}   X        60        6   DZ      _{X        Y}  via   X        50        7   DZ      _{X        Y}   X        50        5   via   via   DY      _{X        Z}   X          60      8   c(X,Y)"

changes" t0   t1   t2   t3continues on…"   Algorithm"

via   DY      _{X        Z}  

X          60      8  

13   via   DY      _{X        Z}   X   DZ      _{X        Y}   X   via  

The Count to Infinity Problem

  The  “poisoned  reverse”  technique  

  If  Z  routes  through  Y  to  get  to  X:  

Then  Z  tells  Y  that  Z’s  distance  to  X  is  infinite    

4 1 50 Z Y X via   DY      _{X        Z}   X   via   DZ      _{X        Y}   X   via   DY      _{X        Z}   X   via   DZ      _{X        Y}   X        50        5   t_-­‐3   t_-­‐2   t_-­‐1   Initialization…" 4        ∞   50      ∞   4        51   50        5   4        ∞   14   DZ      _{X        Y}  via   X   via   DY      _{X        Z}   X   via   DZ      _{X        Y}  

The Count to Infinity Problem

  The  “poisoned  reverse”  technique  

  If  Z  routes  through  Y  to  get  to  X:  

Then  Z  tells  Y  that  Z’s  distance  to  X  is  infinite  

4 1 50 Z Y X 60   via   DY      _{X        Z}   X          4        ∞   via   DZ      _{X        Y}   X        50        5   via   DY      _{X        Z}   X        60      ∞   via   DY      _{X        Z}   X        60      ∞   X        50      61   DZ      _{X        Y}  via   X        50      61   DZ      _{X        Y}   X        50        5   via   c(X,Y)"

changes" t0   t1   t2   t3   terminates"Algorithm"

via   DY      _{X        Z}   X          60    51   t₄   60    51   50      ∞  

  (Will  this  completely  solve  the  problem?)