Multiple path routing: An alternate metric

1
Multiple path routing An alternate metric
Computer Networks - CMPUT 640 Project Presentation

• By
• Frederick Vizeacoumar

2
Agenda

• Introduction
• Motivation
• Previous Works results
• Simulation Environment
• Preliminary Results
• Summary Future Work
• Reference

3
Introduction

• Introduction of integrated services - network are
  being forced to cater - different traffic -
  involve routing protocols
• If the path from source to destination is not
  sufficient, alternate path is next choice
• QoS routing - path based on some criteria
• many QoS routing protocols - introduced

4
Intro...

• some metrics like end-to-end delay depends on the
  requested bandwidth
• some previous researches 2 - translated these
  delays into bandwidth requirement - so many
  studies used bandwidth as the sole metric
• some works also use the metric - number of hops
  as another metric - efficiency and performance of
  the network

5
Motivation

• Many path selection algorithm are available -
  widest shortest path - shortest widest path
• best - widest shortest path
• all concerned about - bandwidth no. of hops
• busy state of the links across the node is not
  concerned

6
Motivation
BW 1
What if we consider the probability of busy state
of the link ?
7
Previous Works

• Many routing algorithm were proposed
• Parallel multi-path routing 3
• synchronization problem - similar to our assn 4
• equal cost multi-path routing2
• not true for all destination
• Link state information should also be an
  important concern
• for simplicity we use on-demand path construction
  scheme

8
Results - Previous works

• Some previous works reveals that many environment
  parameters are to be taken care5
• We mainly focus - Y. Jia, I. Nikoladis,
  P.Gburzynski, Multiple Path routing in Networks
  with Inaccurate Lin State information1 -
  widest shortest path is better - concentrates on
  Link state accuracy - extend some part

9
Simulation Environment

• Topology
• Traffic Model
• Path construction
• Path Selection
• Performance Metric
• Simulation Methodology

10
Assumption

• Certain simulation environment parameters are
  simplified by making some assumptions.
• previous works - prove - link state update scheme
  - not affect the motivation - on demand path
  construction and selection
• Sequential multi-path routing
• Queue at each node stores the request
• compare with widest shortest path algorithm

Where to include the new parameter ?
11
Topology

• We use this ISPs network with 6 nodes, 20 nodes
  and 100 nodes
• link never fail
• from 5 we assume that this topology is important

12
Traffic Model

• Request arrive - Poisson arrival
• Inter-arrival time is generated exponentially
• Bandwidth requirement uniformly distributed bet
  1 5 Mbps
• Duration of holding exponential - mean 3 min.
• Each request receives bandwidth queue at each
  node.
• Destination uniformly distributed

13
Path Construction

• Following 1,
• Node struct NodeId, Bandwidth, Hops,
  probability(usage), parent
• Queue stores node struct priority usage
• Counter maintains no. of path for each
  destination from a given source s
• Pathv(i) - stores the i-th path to destination
  node v
• Bwvi stores max available BW for node v, i-th
  path
• hopvi stores number of intermediate for node
  v, i-th path

14
Path Construction

• Bellman Ford algorithm ignores equal hop
  multiple count 4
• Dijkstras algorithm 1, 4 modify to our
  requirement
• Of nodes 2 3, node 2 has higher bw so node 2 is
  explored

15
Path Construction

• Node 2 have one path
• Dotted line from 2 to 1 is also explored but as
  it makes cycle, we ignore
• Similarly next 4 3,5,6 not 2

16
Path Construction
k 2 N - node n neighbor P path ig ignored

• N 1 n 2 3 / P 1 N 6 n - / P 1-2-4-6 /
  ig 2, 4
• N 2 n 4 6 / P1-2 / ig 1 N 5 n - / P
  1-2-4-3 / ig 3,4
• N 4 n 3, 5 6 / P 1-2-4 / ig 2 N 3 n - /
  P 1-2-4-5-3 / ig 1, 4 5
• N 3 n 5 / P 1-2-4-3 / ig 1, 4 N 6 n 4 / P
  1-2-6 / ig 2
• N 5 n 3 / P 1-2-4-5 / ig 4 N 4 n 3 5 /
  P 1-2-6-4 / ig 2 6
• We dont use the bandwidth but the link busy state

17
Path Selection

• As we have constructed the path based on usage,
  we take the path with max bandwidth capacity that
  could be used.
• Widest Less utilized Shortest path is selected
• To compare the result, we also use Widest
  shortest path

18
Performance Metrics

• We measure the maximum number of times the link
  of the network is utilized.
• We also measure the number of time the request
  was hold in waiting state and sent.
• Time to process certain specific number of
  transmissions request.

19
Simulation Methodology

• Event struct source, dest, bw, time, hold time,
  event type, path and hops
• Event queue stores event priority time
• Type of event
• Request
• Allotted or Hold
• To compare, the usage of links run simulation
  for no. of allotted.

20
Simulation Methodology

• Generate request for each nodes with
• Exponential arrival time
• Uniform bandwidth
• Uniform destination
• Exponential hold time mean 3 mins
• Take the event with least time and move the
  current time counter to the time in event.

21
Simulation Methodology

• If request event,
• Construct path given source residual bandwidth
  of the network after previous allocation
• Select path
• If path available then calculate the residual
  bandwidth, reschedule as allotted event by adding
  the time with req. hold time
• Otherwise reschedule the same request counter
  to count no. of delayed allocation
• After scheduling (if path found)
• Generate another request for the same source
  diff. Destination and diff. Bandwidth.

22
Simulation Methodology

• If allotted event,
• Remove move the event from the event queue and
  release the bandwidth that was used
• Different variables are maintained to calculated
  the number of transmission requested, served and
  currently in the network.
• We also maintain a separate list that reveals how
  many times each link in the network was used.

23
Preliminary Results

• It was found that certain links used max no. of
  times in WSP algorithm is more than WLUSP
  algorithm
• Fixed source destination, BW req. 1, it was
  found that WSP ignores some paths while WLUSP
  utilizes these paths.
• Overall delay to serve certain no. of request is
  more in WLUSP
• Blocked and allotted transmission count is also
  more in WLUSP

24
Summary Future Work

• Using the probability of the links, helps in
  improving network performance
• Delay in serving no. of blocked and allotted
  transmission increases
• This metric good distribute load on the
  network
• Faster and less waiting time WSP
• Possible future work try link state update
  information using update policies

25
References

• Y. Jia, I. Nikoladis, P.Gburzynski, Multiple
  Path routing in Networks with Inaccurate Lin
  State information.
• R.A Guerin, A.Orda and D.Williams QoS Routing
  Mechanisms and OSPF Extentions.
• I.Cidon, R.Rom and Y.Shaviti Multi-Path Routing
  Combined with Resource Reservation.
• G.Apostolopoulos, R Guerin, S Kamat, A.Orda,
  T.Przygienda and D.Williams QoS Routing
  Mechanisms and OSPF Extentions.
• G.Apostolopoulos, R Guerin and S. K. Tripati QoS
  Routing A Performance Perspective.
• E.Q.V. Martins, M.M.B.Pascoal and J.L.E.Santos,
  The K shortest paths problem.

26
Questions