lecture 15: link state routingcseweb.ucsd.edu/.../sp20/cse123-a/lectures/123-sp20-l15.pdfcse 123...

CSE 123: Computer NetworksAaron Schulman

Lecture 15:Link state routing

• Destination-based routing

• Intra-domain routing• Link-state routing

2

Lecture 15 Overview

CSE 123 – Lecture 15: Link-state Routing

! Routingu All addresses are globally known

» No connection setupu Host sends packet with destination address in header

» No path state; only routers need to worry about failureu Distributed routing protocol used to routing tables

! Forwardingu Router looks up destination in table

» Must keep state proportional to destinations rather than connections

u Lookup address, send packet to next-hop link» All packets follow same path to destination

! In Practice this is what we use for IP routing

3

Destination-based Forwarding


! The routing table at A, lists – at a minimum – the next hops for the different destinations

D

G

A

F

E

B

C

Dest Next Hop

B BC CD CE EF FG F

4

Routing Tables


! Essentially a graph theory problemu Network is a directed graph; routers are vertices

! Find “best” path between every pair of verticesu In the simplest case, best path is the shortest path

D

G

A

F

E

B

C =router

=link

X

11

1

11

1

1

11 =cost

5

Routing on a Graph


! How to choose best path?u Defining “best” can be slippery

! How to scale to millions of destinations (addresses)?u Minimize control messages and routing table size

! How to adapt to failures or changes?u Node and link failures, plus message loss

6

Routing Challenges


! Routing within a network/organizationu A single administrative domainu The administrator can set edge costs

! Overall goalsu Provide intra-network connectivityu Adapt quickly to failures or topology changesu Optimize use of network resources

! Non-goalsu Extreme scalabilityu Lying, and/or disagreements about edge costs

7

Intra-domain Routing


! Staticu Type in the right answers and hope they are always trueu …So far

! Link stateu Tell everyone what you know about your neighborsu Today’s lecture!

! Distance vectoru Tell your neighbors what you know about everyoneu Next time…

8

Basic Approaches


! Two phasesu Reliable flooding

» Tell all routers what you know about your local topologyu Path calculation (Dijkstra’s algorithm)

» Each router computes best path over complete network

! Motivationu Global information allows optimal route computationu Straightforward to implement and verify

9

Link-state Routing


! Reliable floodingu Each router transmits a Link State Packet (LSP) on all linksu A neighboring router forwards out all links except incoming

» Keep a copy locally; don’t forward previously-seen LSPs

! Challengesu Packet lossu Out-of-order arrival

! Solutionsu Acknowledgments and retransmissionsu Sequence numbersu Time-to-live for each packet

Broadcasting Link State

10CSE 123 – Lecture 15: Link-state Routing

! LSP generated by X at T=0! Nodes become orange as they receive it

! Entire process repeats with LSPs for A, B, C, …u Actually in runs in parallel

X A

C B D

X A

C B D

X A

C B D

X A

C B D

T=0 T=1

T=2 T=3

11

Flooding Example


Dijkstra’s Shortest Path Tree! So you have all of these LSPs. Now what?! Graph algorithm for single-source shortest path tree

(find best route to all nodes)

S ß {}Q ß <remaining nodes keyed by distance>While Q != {}

u ß extract-min(Q) u = node with lowest costS ß S plus {u}Within Q:

for each node v adjacent to u“relax” the cost of v is it cheaper to go

through u?

ß u is done


Dijkstra Example – Step 1

10

2 3

5

2

1

4 67

90

! Green nodes are “confirmed”

! Yellow nodes are “tentative”

! We can add ourselves to “confirmed”


Example – Step 2

10

2 3

5

2

1

4 67

90

10

5



! First look at neighbors! “5” is cheaper than

“10”! We can confirm path

with cost “5”


Example – Step 3

10

2 3

5

2

1

4 67

90

8

5 7

14



! Update costs! Look at 5’s neighbors! 7 is cheapest! We can confirm path

with cost 7


Example – Step 4

10

2 3

5

2

1

4 67

90

8

5 7

13



! Update costs! 7 has no new

neighbors! 8 is cheapest! We can confirm 8


Example – Step 5

10

2 3

5

2

1

4 67

90

8

5 7

9



! Update costs! No new neighbors! 9 is cheapest! We can confirm path

with cost 9


Example – Done

10

2 3

5

2

1

4 67

90

8

5 7

9


! Need to remove failed/old links from topologyu LSPs carry sequence numbers to distinguish new from oldu Routers only accept (and forward) the “newest” LSPu Send a new LSP with cost infinity to signal a link down

! But also need to remove entire routersu Put a Time To Live (TTL) timer on each LSP u TTL in every LSP, decremented periodically by each routeru When TTL = 0, purge the LSP and flood the network with an

LSP with TTL 0 to tell everyone else to do the same

19

Making Something Disappear


! Triggered by a topology changeu Link or node failure/recovery oru Configuration change like updated link metricu Converges quickly, but can cause flood of updates

! Periodicallyu Typically (say) every 30 minutesu Corrects for possible corruption of the datau Limits the rate of updates, but also failure recovery

When to Flood?


! Getting consistent routing information to all nodesu E.g., all nodes having the same link-state databaseu Until routing protocol converges, strange things happen…

! Consistent forwarding after convergenceu All nodes have the same link-state databaseu All nodes forward packets on shortest pathsu The next router on the path forwards to the next hop

Convergence


! Detection delayu A node does not detect a failed link immediatelyu … and forwards data packets into a black holeu Depends on timeout for detecting lost LSPs

32

2

1

14

1

4

5

3

Transient Disruptions


! Inconsistent link-state databaseu Some routers know about failure before othersu The shortest paths are no longer consistentu Can cause transient forwarding loops

32

2

1

14

1

4

5

3

32

2

11

41

4 3

Transient Disruptions


! Sources of convergence delayu Detection latencyu Flooding of link-state informationu Shortest-path computationu Creating the forwarding table

! Poor performance during convergence periodu Lost packets due to black holes and TTL expiryu Looping packets consuming resourcesu Out-of-order packets reaching the destination

! Very bad for VoIP, online gaming, and video

Convergence Delay


! Faster detectionu Smaller LSP timersu Link-layer technologies that can detect failures

! Faster floodingu Flooding immediatelyu Sending link-state packets with high-priority

! Faster computationu Faster processors on the routersu Incremental Dijkstra’s algorithm

! Faster forwarding-table updateu Data structures supporting incremental updates

Reducing Delay


! OSPF (Open Shortest Path First) and IS-ISu Most widely used intra-domain routing protocolsu Run by almost all ISPs and many large organizations

! Basic link state algorithm plus many features:u Authentication of routing messagesu Extra hierarchy: Partition into routing areas

» “Border” router pretends to be directly connected to all routers in an area (answers for them)

u Load balancing: Multiple equal cost routes

26

Real Link-state Protocols


Link State evaluation ! Strengths

u Loop free as long as LS database’s are consistent» Can have transient routing loops – shouldn’t last long

u Messages are smallu Converges quicklyu Guaranteed to converge

! Weaknessesu Must flood data across entire network (scalability?)u Must maintain state for entire topology (database)


For next time…

• Study for the Midterm• Midterm will by async on Canvas: 1 hour to complete after

you start, and available online for 24 hours• Exam covers everything up to Monday’s lecture• Homework 1 and 2 solutions posted on Piazza and website

• Read Chapter 3.4 in P&D (Distance vector)


lecture 15: link state routingcseweb.ucsd.edu/.../sp20/cse123-a/lectures/123-sp20-l15.pdfcse 123...

Documents