Computer Networks(Routing)Frm Net

download Computer Networks(Routing)Frm Net

of 118

Transcript of Computer Networks(Routing)Frm Net

  • 8/13/2019 Computer Networks(Routing)Frm Net

    1/118

    Computer Networks

    Section1. Routing Algorithms andNetwork Layer Protocol

    Presenter: Shu-Ping Lin

  • 8/13/2019 Computer Networks(Routing)Frm Net

    2/118

    Outline

    Routing Algorithms

    The Network Layer in The Internet

  • 8/13/2019 Computer Networks(Routing)Frm Net

    3/118

    Outline

    Routing Algorithms

    The Network Layer in The Internet

  • 8/13/2019 Computer Networks(Routing)Frm Net

    4/118

    Store-and-Forward Packet Switching

  • 8/13/2019 Computer Networks(Routing)Frm Net

    5/118

    Routing Algorithm

    Routing algorithm

    Datagram

    Virtual circuit

    Differences between routing and forwarding

    Routing versus forwarding

  • 8/13/2019 Computer Networks(Routing)Frm Net

    6/118

    Routing Algorithm (contd)

    Requirement of routing algorithm

    Correctness and Simplicity

    Robustness

    Stability

    Fairness

    Optimality

  • 8/13/2019 Computer Networks(Routing)Frm Net

    7/118

    Datagram Routing

  • 8/13/2019 Computer Networks(Routing)Frm Net

    8/118

    Virtual-Circuit Routing

  • 8/13/2019 Computer Networks(Routing)Frm Net

    9/118

    Comparison

  • 8/13/2019 Computer Networks(Routing)Frm Net

    10/118

  • 8/13/2019 Computer Networks(Routing)Frm Net

    11/118

    Optimality Principle

    If routerJis on the optimal path from routerIto routerK, the the optimal path fromJto

    Kalso falls along the same route. Sink tree

    The set of optimal routes from all sources to agiven destination

    The goal of all routing algorithms is todiscover and use the sink trees for allrouters.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    12/118

    Sink Tree

  • 8/13/2019 Computer Networks(Routing)Frm Net

    13/118

    Shortest Path Routing

    Link metric

    Dijkstra algorithm

  • 8/13/2019 Computer Networks(Routing)Frm Net

    14/118

    Flooding

    Every incoming packet is sent out on everyoutgoing link except the one it arrived on.

    Termination of flooding processHop counter

    Record of packet which has been flooded

    Applications of floodingMilitary application

    Comparison

  • 8/13/2019 Computer Networks(Routing)Frm Net

    15/118

  • 8/13/2019 Computer Networks(Routing)Frm Net

    16/118

    Distance Vector Routing (contd)

  • 8/13/2019 Computer Networks(Routing)Frm Net

    17/118

    Distance Vector Routing (contd)

    Converge to the correct answer, but do so

    slowly.

    It reacts rapidly to good news, but leisurelyto bad news.

    The count-to-infinity problem

  • 8/13/2019 Computer Networks(Routing)Frm Net

    18/118

    Count-to-Infinity Problem

  • 8/13/2019 Computer Networks(Routing)Frm Net

    19/118

    Problem of DVR

    Does not take line bandwidth into account.

    Take too long to converge.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    20/118

    Link State Routing

    Learning about the neighbors

    Sending a special HELLO packet on each

    point-to-point line

    Measuring line cost to each of its neighbors

    Round-trip time

    Traffic loadLine bandwidth

  • 8/13/2019 Computer Networks(Routing)Frm Net

    21/118

    Link State Routing (contd)

    Building link state packet

  • 8/13/2019 Computer Networks(Routing)Frm Net

    22/118

    Link State Routing (contd)

    Distributing the link state packets

    Flooding is used to distribute the link state

    packet.Each packet contains a sequence number that is

    incremented for each new packet sent.

    Computing the new routes Dijkstras algorithm can be run locally to

    construct the shortest path to all destinations.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    23/118

    Hierarchical Routing

    As networks grow in size, the router routing

    tables grow proportionally.

    Each router has responsible for its regionand knows nothing about the internal

    structure of other router.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    24/118

    Hierarchical Routing (contd)

  • 8/13/2019 Computer Networks(Routing)Frm Net

    25/118

    Hierarchical Routing (contd)

    Penalty of increased path length

    The optimal number of levels for an N

    router subnets is ln N.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    26/118

    Broadcasting Routing

    The source simply send a distinct packet to

    each destination.

    Waste bandwidth

    Have to know complete list of all destinations

    Flooding

    Generate too many packets

    Consume too much bandwidth

  • 8/13/2019 Computer Networks(Routing)Frm Net

    27/118

    Broadcasting Routing (contd)

    Multidestination routing

    Each packet contains a list of destinations.

    The destination set is partitioned among theoutput lines.

    Using spanning tree

    Routers must know which of its lines belong tothe spanning tree in advance.

    Copy an incoming packet onto all the spanninglines except the on it arrived on.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    28/118

    Broadcasting Routing (contd)

    Reverse path forwarding

    Router checks to see if the packet arrived on the

    line that is normally used for sending packets tothe source of the broadcast.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    29/118

    Multicast Routing

    Each router computes a spanning tree

    covering all other routers.

    When a process sends a multicast packet toa group

    Examining the spanning tree of this group

    Removing all lines that do not lead to hosts thatare members of group

  • 8/13/2019 Computer Networks(Routing)Frm Net

    30/118

    Multicast Routing (contd)

  • 8/13/2019 Computer Networks(Routing)Frm Net

    31/118

    Multicast Routing (contd)

    Pruning spanning tree

    Reverse path forwarding

    Router with no hosts interested in a particulargroup sends a PRUNE message.

    The subnet is recursively pruned.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    32/118

    Multicast Routing (contd)

    Disadvantage of algorithm, suppose thatnetwork has ngroups, each with an average

    of mmembers.For each group, mpruned spanning trees must

    be stored

    Total of mntrees

    Core-based treeA host wanting to multicast sends packets to the

    core, which does the multicast along thespanning tree.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    33/118

    Routing for Mobile Hosts

  • 8/13/2019 Computer Networks(Routing)Frm Net

    34/118

    Routing for Mobile Hosts (contd)

    Mobile host

    Migratory hosts

    Roaming hosts All hosts are assumed to have a permanent

    home locationthat never changes.

    Hosts also have a permanent home addressthat can be used to determine their homelocation.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    35/118

    Routing for Mobile Hosts (contd)

    Each area has a home agentwhich keeps

    track of hosts whose home is in the area, but

    who are currently visiting another area. Each area has foreign agentwhich keeps

    track of all mobile hosts visiting this area.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    36/118

    Routing for Mobile Hosts (contd)

    Registration procedure

    Foreign agent searching

    Registration

    Foreign agent contacts the mobile hosts home

    agent.

    Verification of mobile hosts home agent Acknowledgement from the mobile hosts

    home agent

  • 8/13/2019 Computer Networks(Routing)Frm Net

    37/118

  • 8/13/2019 Computer Networks(Routing)Frm Net

    38/118

  • 8/13/2019 Computer Networks(Routing)Frm Net

    39/118

    Routing in Ad Hoc Networks

    AODV (Ad hoc On-demand Distance

    Vector)

    Distant relative of the Bellman-Ford distancevector algorithm

    Considering the limited bandwidth and low

    battery life

    On-demand algorithm

  • 8/13/2019 Computer Networks(Routing)Frm Net

    40/118

    Routing in Ad Hoc Networks (contd)

    AODV algorithm maintains a table at each

    node, keyed by destination and which

    neighbor to send packets in order to reachdestination.

    Route request packet

  • 8/13/2019 Computer Networks(Routing)Frm Net

    41/118

  • 8/13/2019 Computer Networks(Routing)Frm Net

    42/118

  • 8/13/2019 Computer Networks(Routing)Frm Net

    43/118

    Routing in Ad Hoc Networks (contd)

    Destination constructs a ROUTE REPLYpacket which follows the reverse path to

    source. Each intermediate node enters this packet

    into local routing table when

    No route to destination is known.

    Sequence number for destination in REPLY isgreater than the value in the routing table.

    Sequence number is equal but the new route isshorter.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    44/118

    Routing in Ad Hoc Networks (contd)

    Problem of mobility

    Route maintenance

    Periodically, each node broadcasts aHello

    message.

    If no response is returned, router prune this link.

    Active neighbor

  • 8/13/2019 Computer Networks(Routing)Frm Net

    45/118

    Routing in Ad Hoc Networks (contd)

  • 8/13/2019 Computer Networks(Routing)Frm Net

    46/118

    Routing in Ad Hoc Networks (contd)

    Critical difference between AODV and

    DVR

    Nodes do not send out periodic broadcastscontaining their entire routing table.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    47/118

  • 8/13/2019 Computer Networks(Routing)Frm Net

    48/118

    Node Lookup in Peer-to-Peer Networks

    (contd) Finger table

    Start field = k +2i(modulo 2m)

    If keyfalls between kandsuccessort (k),thenthe node holding information about keyis

    successor (k).

    Otherwise, the entry whosestart field is the

    closest predecessor of key is tried.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    49/118

    Node Lookup in Peer-to-Peer Networks

    (contd)

  • 8/13/2019 Computer Networks(Routing)Frm Net

    50/118

    Node Lookup in Peer-to-Peer Networks

    (contd) Noderjoining

    New node raskssuccessor (r) for its predecessor.

    Insert rin between successor and predecessor. Successor should hand over those keys in the range

    predecessor (r)-r, which now belong to r.

    Every node runs a background process that

    periodically recomputes each finger by callingsuccessor.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    51/118

    Outline

    Routing Algorithm

    The Network Layer in The Internet

  • 8/13/2019 Computer Networks(Routing)Frm Net

    52/118

    The Network Layer in The Internet

    Principles for network design

    Make sure it works.

    Keep it simple.Make clear choices.

    Exploit modularity.

    Expect heterogeneity.

    Avoid static options and parameters.

    Look for a good design; it need not be perfect.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    53/118

    The Network Layer in The Internet

    (contd)Think about scalability.

    Consider performance and cost.

    Network layer provides best-efforts way totransport datagrams from source to

    destination.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    54/118

    The IP Protocol

    The IPv4 header

  • 8/13/2019 Computer Networks(Routing)Frm Net

    55/118

    IP Address

    IP address formats

  • 8/13/2019 Computer Networks(Routing)Frm Net

    56/118

    IP Address (contd)

    Special IP address

  • 8/13/2019 Computer Networks(Routing)Frm Net

    57/118

    Subnet

    Split a network into several parts for

    internal use.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    58/118

    Subnet (contd)

    Some bits are taken away from the hostnumber to create a subnet number.

    For example, a university can use a 6-bitsubnet number and a 10-bit host number,allowing fro up to 64 subnets.

    Outside the network, the subnet is notvisible, so allocating a new subnet does notrequire contacting ICANN.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    59/118

    Subnet (contd)

    Subnet mask indicates the split between

    network + subnet number and host.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    60/118

  • 8/13/2019 Computer Networks(Routing)Frm Net

    61/118

    Network Address Translation (contd)

  • 8/13/2019 Computer Networks(Routing)Frm Net

    62/118

    Network Address Translation (contd)

    Use TCP port to distinguish the traffic.

    TCPsource portfield is replaced by an

    index into the NAT boxs translation table.

    Each entry contains original IP address and

    original source port.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    63/118

    Network Address Translation (contd)

    Violation of the architectural model of IP

    Every IP address uniquely identifies a machine.

    NAT changes the Internet from aconnectionless network to a kind ofconnection-oriented network.

    Violation of protocol layering

    layer kmay not make any assumptions aboutwhat layer k+1has put into the payload field.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    64/118

    Internet Control Protocols

    Internet Control Message Protocol (ICMP)

  • 8/13/2019 Computer Networks(Routing)Frm Net

    65/118

    Address Resolution Protocol

    How do IP address get mapped onto data

    link layer addresses, such as Ethernet?

    Configuration file ARP

    Broadcast a packet onto the Ethernet asking:

    Who owns IP address xxx.xxx.xxx.xxx.Correspondent will reply with its Ethernet

    address

  • 8/13/2019 Computer Networks(Routing)Frm Net

    66/118

    Address Resolution Protocol (contd)

  • 8/13/2019 Computer Networks(Routing)Frm Net

    67/118

  • 8/13/2019 Computer Networks(Routing)Frm Net

    68/118

    RARP, BOOTP, and DHCP

    When a computer is booted how does it get

    the IP address.

    RARPBroadcast a packet to ask RARP server.

    Packet cannot be forwarded by routers, so

    RARP server is needed on each network.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    69/118

  • 8/13/2019 Computer Networks(Routing)Frm Net

    70/118

    RARP, BOOTP, and DHCP (contd)

    DHCP

  • 8/13/2019 Computer Networks(Routing)Frm Net

    71/118

    OSPFThe Interior Gateway Routing

    Protocol

    Routing algorithm within as autonomous

    system (AS) is called an interior gateway

    protocol. The original Internet interior gateway

    protocol was distance vector protocol (RIP).

    Link state protocol replaced RIP in 1979. OSPF began work in 1988.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    72/118

    OSPFThe Interior Gateway Routing

    Protocol (contd)

    Internet is made up of a large number ofautonomous systems.

    Each AS is operated by a differentorganization and can use its own routingalgorithm.

    Every AS has a backbonearea.

    All areas are connected to the backbone andcan communicate each other via backbone.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    73/118

  • 8/13/2019 Computer Networks(Routing)Frm Net

    74/118

    OSPFThe Interior Gateway Routing

    Protocol (contd)

    Using flooding, each router informs all theother router in its area of its neighbors andcosts.

    This information allows each router toconstruct the graph for its area and computethe shortest path.

    Backbone routers accept information fromthe area border routers in order to computethe best route from each backbone router toevery other router.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    75/118

    BGPThe Exterior Gateway Routing

    Protocol

    All interior gateway protocol has to do is to move

    packets as efficiently as possible without worrying

    about politics.

    Exterior gateway protocol routers have to worry

    about politics.

    No transit traffic through certain ASes.

    Traffic starting or ending at IBM should not transitMicrosoft.

    Policies are typically manually configured into

    each BGP router and are not part of protocol.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    76/118

    BGPThe Exterior Gateway Routing

    Protocol (contd)

    Stub networks

    Has only one connection to BGP graph andcannot be used for transit traffic.

    Multiconnected networks

    Could be used for transit traffic, except thatthey refuse

    Transit networksWilling to handle third-party packet

  • 8/13/2019 Computer Networks(Routing)Frm Net

    77/118

    BGPThe Exterior Gateway Routing

    Protocol (contd)

    Border Gateway Protocol (BGP)

    Fundamentally a distance vector protocol, but

    each BGP router keeps track of the path used.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    78/118

    Internet Multicasting

    IP uses class D address to support multicast.

    Two kinds of group address

    Permanent address

    Temporary address

  • 8/13/2019 Computer Networks(Routing)Frm Net

    79/118

    Internet Multicasting (contd)

    224.0.0.1 All systems on a LAN

    224.0.0.2 All routers on a LAN

    224.0.0.5 All OSPF routers on a LAN

    224.0.0.6 All designated OSPF routers on a

    LAN

    Temporary groups must be created before

    they can be used.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    80/118

    Internet Multicasting (contd)

    Multicasting routing is done using spanning

    tree.

    Each multicast router exchangesinformation with its neighbors, using a

    modified distance vector protocol.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    81/118

    Mobile IP

    Increment of portable computers.

    Major goals

    Mobile host must be able to use its home IPaddress anywhere.

    Changes to software and router are not

    permitted.

    No overhead should be incurred when a mobile

    host is at home.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    82/118

  • 8/13/2019 Computer Networks(Routing)Frm Net

    83/118

  • 8/13/2019 Computer Networks(Routing)Frm Net

    84/118

    IPv6

    While NAT may buy a few more yearstime, IP in its current form (IPv4) isnumbered.

    IETF issued a call for proposal anddiscussion in RFC 1550.

    IPv6 is not compatible with IPv4, but it is

    compatible with other auxiliary Internetprotocol.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    85/118

    IPv6 (contd)

    Main features

    Longer addresses

    Simplification of the headerBetter support for options

    Security

    Quality of service

  • 8/13/2019 Computer Networks(Routing)Frm Net

    86/118

    IPv6 (contd)

    The IPv6 fixed header

  • 8/13/2019 Computer Networks(Routing)Frm Net

    87/118

    IPv6 (contd)

    New notation of 16-byte address

    8000:0000:0000:0000:0123:4567:89AB:CDEF

    8000::123:4567:89AB:CDEF IPv4 can be written as ::192.168.20.46

    Vanishment of IPv4 headers

    IHL

    Protocol

    Checksum

  • 8/13/2019 Computer Networks(Routing)Frm Net

    88/118

    IPv6 (contd)

    Extension header is encoded as (Type,

    Length, Value) tuple.

    Type is a 1-byte field telling which option thisis.

    Length is a 1-byte field telling how long the

    value is.

    Value is any information required.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    89/118

    IPv6 (contd)

    IPv6 extension header

  • 8/13/2019 Computer Networks(Routing)Frm Net

    90/118

    IPv6 (contd)

    Controversies

    Hop limit field length

    Maximum packet sizeChecksum

    Mobile issue

  • 8/13/2019 Computer Networks(Routing)Frm Net

    91/118

    Section2. The Internet Transport Protocols:TCP

  • 8/13/2019 Computer Networks(Routing)Frm Net

    92/118

    Introduction to TCP

    Function of providing reliable end-to-end

    byte stream over an unreliable internetwork.

    Lack of IPNo guarantee that packets will be delivered

    properly.

    Packets may arrive in the wrong order.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    93/118

    TCP Service Model

    TCP service is obtained by both sender and

    receiver creating sockets.

    Each socket has IP address and a 16-bitnumber called port.

    Port number below 1024 are called well-

    known portsand reserved for standardservice.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    94/118

    TCP Service Model (contd)

    Well-Known Ports

  • 8/13/2019 Computer Networks(Routing)Frm Net

    95/118

    TCP Service Model (contd)

    Characteristics of TCP connections

    Full duplex

    Point-to-pointByte stream, not a message stream

  • 8/13/2019 Computer Networks(Routing)Frm Net

    96/118

    TCP Service Model (contd)

    A key feature of TCP is that every packet

    on a TCP connection has its own 32-bit

    sequence number. TCP segmentconsisting of fixed 20-byte

    header is used for exchanging data between

    sender and receiver.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    97/118

    TCP Service Model (contd)

    Segment size issue

    Fit in the 65515-byte IP paylo.ad

    Cant exceed maximum transfer unit (MTU).

    MTU is generally 1500 bytes.

    The basic protocol used by TCP is thesliding window protocol.

    Timer

    Acknowledgement number

  • 8/13/2019 Computer Networks(Routing)Frm Net

    98/118

    TCP Segment Header

  • 8/13/2019 Computer Networks(Routing)Frm Net

    99/118

    TCP Connection Establishment

    Three-way handshake.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    100/118

    TCP Connection Release

    To release a connection , either party can

    send a TCP segment with theFINbit set. When the FINis acknowledged, that

    direction is shut down for new data.

    Normally, four TCP segments are required

    to release a connection.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    101/118

  • 8/13/2019 Computer Networks(Routing)Frm Net

    102/118

    TCP Transmission Policy

  • 8/13/2019 Computer Networks(Routing)Frm Net

    103/118

    TCP Transmission Policy (contd)

    When the window size is 0, the sender stop

    sending data to receiver except

    Urgent data may be sent.Send a 1-byte segment to make the receiver

    reannounce the next byte expected and window

    size.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    104/118

    TCP Transmission Policy (contd)

    Consider the worst case in performance

    issue.

    For receiversDelay acknowledgements and window updates

    for 500 msec in the hope of acquiring some

    data.

    For senders

    Nagles algorithm

  • 8/13/2019 Computer Networks(Routing)Frm Net

    105/118

    TCP Transmission Policy (contd)

    Silly window syndrome

    Data are passed to the sending TCP entity in

    large blocks, but an interactive application onthe receiving side reads data 1 byte at a time.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    106/118

    TCP Transmission Policy (contd)

    Clarks algorithm

    It forces window update to wait until it has a

    decent amount of space available. Combination of Nagles and Clarks

    algorithm.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    107/118

    TCP Congestion Control

    When the load offered to any network ismore than it can handle, congestion buildsup.

    TCP achieves congestion control bydynamically manipulating the window size.

    First step of congestion control: detection.

    All the Internet TCP algorithms assume thattimeouts are caused by congestion.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    108/118

    TCP Congestion Control (contd)

  • 8/13/2019 Computer Networks(Routing)Frm Net

    109/118

    TCP Congestion Control (contd)

    Two potential problemsnetwork capacity

    and receiver capacity.

    Each sender maintains two windowsreceiver window and congestion window.

    The number of bytes that may be sent is the

    minimum of the two windows.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    110/118

    TCP Congestion Control (contd)

    Slow start

  • 8/13/2019 Computer Networks(Routing)Frm Net

    111/118

    TCP Timer Management

    Retransmission timer

    Difficulties of setting retransmission timer

  • 8/13/2019 Computer Networks(Routing)Frm Net

    112/118

  • 8/13/2019 Computer Networks(Routing)Frm Net

    113/118

    TCP Timer Management (contd)

    Jacobson proposed proposed to use mean

    deviationas a cheap estimator of the

    standard deviation. ||)1( MRTTDD

    DRTTTimeout 4

  • 8/13/2019 Computer Networks(Routing)Frm Net

    114/118

    TCP Timer Management (contd)

    Potential problem

    When the ack of retransmission packet comes

    in, it is unclear whether the ack refers to thefirst transmission or a later one.

    Karns algorithm

    Persistence timer is design to prevent the

    deadlock.

    Keepalive timer.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    115/118

    Wireless TCP and UDP

    Problem due to congestion control.

    TCP assumes that timeouts are caused by

    congestion, not by lost packets. A packet is lost on a wired network, the

    sender should slow down.

    When one is lost on a wireless network, thesender should try harder.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    116/118

    Wireless TCP and UDP (contd)

    Bakne and Badrinath propose indirect TCP

    which split the TCP connection into two

    separate connections.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    117/118

    Wireless TCP and UDP (contd)

    The advantage of this scheme is that

    bothconnections are now homogeneous.

    Timeouts on the first connection can slowthe sender down, whereas timeouts on the

    second one can speed it up.z

    The disadvantage of the scheme is that itviolates the semantics of TCP.

  • 8/13/2019 Computer Networks(Routing)Frm Net

    118/118

    Transactional TCP

    Efficiency problem of TCP