ChannelSharing and Multicasting - PowerPoint PPT Presentation

1 / 45
About This Presentation
Title:

ChannelSharing and Multicasting

Description:

Apply to the generalized shuffle-exchange-based multihop architecture, called GEMNET ... a generalization of the shuffle-exchange connectivity pattern using ... – PowerPoint PPT presentation

Number of Views:26
Avg rating:3.0/5.0
Slides: 46
Provided by: deronCsi
Category:

less

Transcript and Presenter's Notes

Title: ChannelSharing and Multicasting


1
Chapter 7
  • Channel-Sharing and Multicasting

2
Outline
  • 7.1 Introduction
  • 7.2 Background
  • 7.3 Shared-Channel Multi-hop GEMNET
  • 7.4 Performance Evaluation
  • 7.5 Illustrative Example Unicast Traffic
  • 7.6 Illustrative Example Multicast Traffic

3
Introduction
  • W (the number of wavelengths) ?N (the number of
    nodes)
  • A general method using channel-sharing is used to
    construct practical multihop networks under this
    limitation.
  • Channel-sharing may be achieved through
    time-division multiplexing (TDM).
  • Apply to the generalized shuffle-exchange-based
    multihop architecture, called GEMNET

4
Multicast
  • Multicasting the ability to transmit information
    from a single source node to multiple destination
    nodes
  • Application
  • (1) multiperson conferences,
  • (2) video lectures where a single speaker can
    address a large audience,
  • (3) "multipoint-LAN interconnection" which allows
    large companies to treat their many
    geographically distributed LANs as a single large
    network, and
  • (4) Collaborative systems. Multicasting is also
    required to make updates in replicated and
    distributed databases.
  • (5) Multiprocessor systems also require
    multicasting for cache coherency and message
    passing.
  • (6) make updates in replicated and distributed
    databases.

5
Introduction
  • For the case w lt N, each of the w channels will
    be assumed to be shared by one or more nodes in a
    time-division multiplexed (TDM) fashion.
  • Each node is equipped with only a single
    transmitter-receiver pair.
  • WDM channels are shared by nodal transmitters
    and receivers in TDM fashion, exhibit an
    interesting and anomalous delay behavior when the
    number of wavelengths is varied.

6
Delay Behavior
7
Multicast on channel-sharing
  • Interestingly, channel-sharing approaches are
    very well suited for multicasting applications in
    a local lightwave network.
  • This is because our methods will entail sharing
    at both the transmitting and the receiving ends.
  • The latter implies that many nodes will be able
    to simultaneously receive the same information
    from a single transmission, thereby increasing
    the receiver through-put (informally defined as
    the average rate of information received by nodes
    in the network) for multi-destination traffic.

8
7.2 Background of Channel-sharing
  • The first extensive treatment of the multihop
    approach in WDM optical networks was provided in
    Acam87.
  • ShuflieNet logical topology,
  • P transmitter-receiver pairs.
  • each node has a degree P, and
  • networks of size N KPK
  • the number of channels w KPK1, where K is the
    number of columns in the ShuffleNet.
  • The idea of channel-sharing is achieved through
    the use of TDM all receivers in a common row of
    the ShuffleNet are assigned the same wavelength.

9
?1
?1
?5
?5
10
Channel-sharing
  • these transmitters must take turns (e.g., in TDM
    fashion) to use this channel.
  • Any packet of information that these two
    transmitters put onto the channel will be "heard"
    by both nodes 2 and 3.
  • The limitation of this strategy is that each node
    requires two transmitter-receiver pairs even in
    the shared case.

11
Multicasting
  • Single-hop System
  • In both BoMu95b and RoAm94, it was shown
    that, for systems with tunable receivers, the
    receiver throughput BoMu95b defined to be the
    number of packets that are delivered in a
    single-transmission slot to various destinations
    in a passive-star network can be greater than
    the transmitter throughput.
  • It was also shown in RoAm94 that receiver
    throughput can be much larger than transmitter
    throughput even in systems with tunable
    transmitters provided w lt N channels are used.

12
Multicast
  • Multihop System
  • FT-FR Not only cost effective, but can also
    provode a good basis for comparison with single
    hop system.
  • WltN, Nkw
  • GEMNET N need not be equal to KPK.
  • Analysis feature
  • Channel-sharing, TDM, fixed length packets,
    M/D/1.
  • Nonzero propagation delay are considered.
  • The effect of channel-sharing on the mean total
    delay experienced by a packet in the network can
    be quantified.
  • The effect of multicast traffic.

13
7.3 Shared-Channel Multihop GEMNET
  • A (K, M, P) GEMNET,
  • N nodes
  • each of degree P
  • arranged in a cylinder of K columns and M nodes
    per column so that nodes in adjacent columns are
    arranged according to a generalization of the
    shuffle-exchange connectivity pattern using
    directed links.
  • Without channel-sharing, the number of channels
    required in a (K, M, P) GEMNET equals NPKMP.

14
Shared Channel Multihop GEMNET
  • When each node can have only one fixed
    transmitter-receiver pair and w lt N, the degree
    of the network takes on a new meaning.
    Channel-sharing through TDM implies a higher
    logical nodal degree (viz., ability to reach more
    nodes directly), albeit with a lower capacity on
    each of the logical links.
  • The nodal degree is redefined as a logical
    quantity P, where P N/w.
  • For analytical convenience, N is assumed to be an
    integral multiple of w to ensure equal and fair
    sharing by all logical links.
  • Thus, we define a (K, M, P) SC_GEMNET
    (Shared-Channel GEMNET) as a GEMNET having K
    columns and M rows, where M is an integral
    multiple, n, of P
  • (i.e., M nP) and K N/M N/(nP).

15
N12, GEMNET
K12, M1
w6
16
W4
w4
17
N12, SC_MEMNET
w3
18
w2
19
average hop distance
  • Diameter D
  • The upper bound of average hop distance h
  • The lower bound of average hop distance h

20
  • In general, two-column GEMNETs (i.e., K 2) have
    the best performance in terms of the average hop
    distance.
  • Therefore, in constructing logical topologies for
    the multihop WDM network, it is, in general, best
    to consider K 2 wherever possible and use the
    next higher value of K as dictated by the
    requirement K N/M N/(nP) w/n, where n is an
    integer which equals the number of channels
    required to provide connectivity between two
    successive column.

21
7.4 Performance Evaluation
  • Derive an expression for the average delay
    encountered by a packet before it is delivered to
    all of its destinations.
  • Assume that communication is time-slotted with a
    slot being equal to the transmission time of a
    packet.
  • For example, if a 1-Gbps channel is used with a
    53-byte packet (or cell) as defined in the ATM
    standard, then the slot duration is 0.424µs.
  • All other delays in the system are normalized to
    this slot duration.
  • Intermediate nodes are capable of examining the
    headers of the packets they receive to determine
    whether the local node is a destination.

22
Notations
  • N Number of nodes in the system.
  • w Number of channels in the system. w is chosen
    such that N/w is an integer.
  • R Round-trip propagation delay between a node
    and the passive star. We assume that all nodes
    are equidistant from the passive star. If needed,
    delay lines can be installed so that R is the
    same for all nodes.
  • F The frame length is equal to N/w. Since TDM is
    used on each of the channels and N/w is the
    number of nodes which share a single channel, a
    frame comprises F slots in which different nodes
    get to transmit.
  • m Multicast destination set size. All packets in
    the network are multicast packets, and each one
    of them has a destination set size specified by a
    constant m.
  • hm The average number of transmissions required
    to deliver information to the
  • gm The average number of hops separating the
    source from each one of the destinations of a
    multicast packet.

23
(No Transcript)
24
(No Transcript)
25
(No Transcript)
26
(No Transcript)
27
7.5 Illustrative Example Unicast Traffic
  • N12 nodes
  • Round-trip delay R2.
  • Load ?0.05 packets per slot per node.
  • Packet size 53bytes (ATM cell).
  • Each channel operates on 100Mbps
  • R2(4.24)8.48µs
  • Node-to-star distance lt1km
  • Load of each node 5Mbps, total 60Mbps.

28
Various Delay
29
Delay Behavior vs. Variation in ?
30
Optimal of wavelength
31
Delay Behavior in Larger Network N120
32
(No Transcript)
33
R0
R1
34
R100
R10
35
(No Transcript)
36
  • Larger network do not always requires an equal
    larger number of wavelength.

37
(No Transcript)
38
7.6 Multicast traffic
  • System A
  • 39-node
  • a single transmitter-receiver pair
  • 39 dedicated channels
  • node degree 1.
  • the entire bandwidth of a channel is available
    to each node.
  • logical topology ring.
  • System B
  • N39, 3 transmitter-receiver pairs
  • 117 dedicated channels node degree 3,
  • the bandwidth available to each node is 3 times
    the bandwidth of a channel.
  • Topology (13, 3, 3)-GEMNET
  • System C
  • 39-node
  • a single transmitter-receiver pair
  • 13 dedicated channels (shared)
  • node degree 3.
  • the entire bandwidth of a channel is available
    to each node.
  • Topology (13, 3, 3)-GEMNET.

39
  • System D
  • A 42-node ring network,
  • employing a single transmitter-receiver pair at
    each node and
  • 42 channels.
  • System E
  • A (7, 6, 6)-GEMNET, employing six
    transmitter-receiver pairs
  • at each node and 42 x 6 252 channels,
  • nodal degree of six.
  • System F
  • A (7, 6, 6)-SC_GEMNET, employing a single
    transmitter-receiver pair at each node and
  • seven channels, leading to a nodal degree of six.
  • System G
  • A (1, 42, 42)-SC_GEMNET, employing a single
    transmitter-receiver pair at each node and one
    channel.
  • This is the classical TDM system.

40
  • We assume that the traffic arrival rate A 0.01
    packets per slot per node,
  • propagation delay R 10 ( node-to-star distance
    of about 1 km,)
  • Under these conditions, we evaluated the delay
    for the various systems for different values of
    destination set size m.

41
(No Transcript)
42
A, B, and C
  • System A with 39 dedicated channels has the worst
    performance.
  • because the only network that can be logically
    implemented for this case is a ring network,
    which can have significant delays due to
    increased hop distance.
  • Our shared-channel scheme with only 13 channels
    and a degree of three performs as well as the
    117-channel system also with degree three.
  • In shared-channel case, it uses fewer channels
    and/or transceivers than the dedicated-channel
    cases (Systems A and B) however, it does perform
    better in terms of delay.

43
(No Transcript)
44
(No Transcript)
45
(No Transcript)
Write a Comment
User Comments (0)
About PowerShow.com