Simulation Design for XCAST Handoff

1
Simulation Design for XCAST Handoff

• Yi Pan
• Yuji Imai
• Meejeong Lee

2
Purpose of Simulation

• To find out the parameters of the proposed
  handoff scheme and their impact on the
  performance
• Performance Comparisons with other schemes
• Existing single binding handoff scheme
• With data link level handoff support
• Without data link level handoff support
• Mobility support using Internet Standard Multicast

3
Performance metric

• Metric to measure the quality of service
• Blackout time
• Handoff time
• Throughput
• Loss rate
• Metric to measure overhead
• Transmission overhead
• Processing overhead
• Storage overhead

4
Performance Metrics

• Blackout time
• Time period during which there is no path to
  receive data from CN
• Measuring method
• Blackout starts when MN finds out loss of links
  to all of the BSs corresponding to the registered
  COAs
• Blackout ends when MN receives an Ack. for any
  COA registration and the BS corresponding to that
  COA is reachable

5
Performance Metrics

• Uncontrolled Traffic
• Traffic sent out to the new link while the sender
  still assumes the traffic is going through the
  same link.
• Measuring method
• Count the number of packets from the new link at
  the mobile node before the first RR arrives the
  sender indicating the new link.

6
Performance Metrics

• Uncontrolled Traffic
• Measuring method (Contd)
• For single binding and XCAST schemes
• any traffic directed to the new link before the
  sender receives the binding update is measured as
  uncontrolled traffic.
• For ISM scheme
• any traffic directed to the new links before the
  sender receives the first Receiver Report
  reporting loss is measured as uncontrolled
  traffic.

7
Performance Metrics

• Measuring method (Contd)
• For XCAST, handoff time starts when there is no
  existing link receiving packets or MN starts to
  receive packet from the new link, whichever comes
  first, and ends when CN detects end of slow start
  on the new link.
• For ISM, handoff time starts when there is no
  existing link receiving packets or MN starts to
  receive packet from the new link, whichever comes
  first, and ends when the throughput equals the
  available bandwidth in the new link.

8
Performance Metrics

• Throughput of traffic
• Average transmission rate for the whole
  simulation period.
• Measuring method
• MN counts the number of received packets except
  for the duplicate ones

9
Performance Metrics

• Loss rate of traffic
• Represent the quality (wholeness) of received
  data
• Measuring method
• CN counts the number of data packets sent out,
  excluding duplicate ones (K)
• MN counts the number of data packets for which
  not a single copy is received (L)
• L/K

10
Performance Metrics

• Overhead measurement
• For the wireless domain where the MN belongs,
  measure the actual overhead in the simulation
• End system outside the wireless domain need to
  measure the overhead it detects.
• The amount of overhead in the rest of the
  Internet is constant and not affected by the
  mobility of the MN for all the schemes

11
Performance Metrics

• Transmission overhead in wireless domain
• Control overhead
• Control messages
• Extra header information
• Data traffic overhead
• Redundant traffic
• Useless data traffic
• Each entity in the wireless domain counts the
  number of control or data packets received

12
Performance Metrics

• Processing overhead in wireless domain
• Control message handling overhead count the
  number of control messages received whenever the
  received control message requires processing at
  the receiving entity
• For each received control message, count the
  number of each specific operation that is
  incurred by that control message

13
Performance Metrics

• Storage overhead in wireless domain
• Mainly, the additional storage size in the
  intermediate routers for each mobile node is
  considered.
• Second, the size of storage in the base station
  is considered.
• Third, the size of storage in the end systems is
  considered.

14
Performance Metrics

• Transmission overhead for Single COA binding
  scheme
• Control overhead
• Binding update message each router in the
  wireless domain counts the number of binding
  update messages received
• Fast mobile IP handoff control message BS counts
  the number of fast mobile IP handoff request
  messages received

15
Performance Metrics

• Transmission overhead for Single COA binding
  scheme
• Data traffic overhead
• Useless traffic received at the old link that can
  not be delivered/buffered measured by the BSs
• number of non duplicate packets received at MN/
  sum of the number of received packets at each
  entity in the wireless domain

16
Performance Metrics

• Processing overhead of single COA binding scheme
• Binding update at HA and CN HA and CN counts the
  number of binding updates received
• Transferring of buffered packets from an old BS
  to a new BS BS counts the number fast mobile IP
  handoff requests received as well as the amount
  of transferred packets

17
Performance Metrics

• Transmission overhead for XCAST handoff
• Control overhead
• Multiple COA binding updates each router in the
  wireless domain
• XCAST header XCAST tunneling routers
• Data traffic overhead
• Redundant traffic duplicated packets received at
  the MN
• Useless traffic BS counts the number of arriving
  packets that can not be delivered to MN
• Number of non duplicate packets received at MN/
  sum of the number of received packets at each
  entity in the wireless domain

18
Performance Metrics

• Processing overhead for XCAST handoff
• HA and CN counts the number of COA updates
• XCAST tunneling nodes counts the number of
  received data packet with XCAST header

19
Performance Metrics

• Transmission overhead for ISM
• Control overhead
• IGMP message(periodically refresh group
  membership from end-systems to base stations)
  counted at each BS
• MOSPF message(group membership advertisement to
  all multicast routers in the network
  periodically) counted at each intermediate
  routers
• Binding update message(periodical update of
  multicast group address binding to HA and CN)
  counted at each router in the wireless domain

20
Performance Metrics

• Transmission overhead for ISM (contd)
• Data traffic overhead
• Redundant traffic duplicated packets received at
  the MN
• Useless traffic BS counts the number of arriving
  packets that can not be delivered to MN
• number of non duplicate packets received at MN/
  sum of the number of received packets at each
  entity in the wireless domain

21
Performance Metrics

• Processing overhead for ISM
• Membership update advertisement triggers
  re-computation of MOSPF group membership
  information each router counts the number of
  valid membership update advertisement received
• The first packet after a membership update
  triggers source-based multicast tree computation
  each router counts the number of first data
  packet received after each membership update.
• HA and CN count the number of binding updates
  received

22
Performance Metrics

• Storage overhead for single binding scheme
• An entry for each existing host is maintained in
  the base station.
• An entry for each previous host requesting
  forwarding service is maintained in the base
  station.
• Only for fast handoff in mobile IP
• An entry for the binding of COA is needed in HA
  and CN.

23
Performance Metrics

• Storage overhead for XCAST scheme
• An entry for each existing host is maintained in
  the base station.
• Note here each host can have multiple base
  stations keeping the entries for it.
• An entry per mobile IP binding is needed in the
  MN.
• An entry contains multiple mobile IP bindings is
  needed in HA and CN.

24
Performance Metrics

• Storage overhead for ISM mobility support
• An multicast entry for each mobile node in all
  the intermediate routers on each source-base
  multicast tree.
• An entry in all intermediate routers in the
  domain for group membership information.
• An entry for each existing host in the base
  stations.
• An entry for each interface attending the
  multicast group is needed in the MN.
• An entry to bind the multicast address to the
  MNs permanent address is needed in HA and CN.

25
Performance Metrics

• Analysis of fixed constant for overhead in
  Internet
• Fixed portion of the overhead for single binding
  scheme
• Fixed portion of the overhead for XCAST handoff
  scheme
• Fixed portion of the overhead for ISM

26
Performance Metrics

• Fixed portion of the overhead for single binding
  scheme
• Major observations of Internet portion
• Routes that the control information and data
  traffic flow through in the Internet are not
  changed.
• Volume of control traffic is not amplified by the
  Internet.

27
Performance Metrics

• Fixed portion of the overhead for single binding
  scheme (Contd)
• Based on the observation, we conclude that we
  have a constant number C1 of forwarding nodes in
  Internet. So the overheads can be analyzed as
• Transmission overhead
• Only binding update messages will go out of
  wireless domain.
• Conceptual formula for the total overhead traffic
  is
• Ctransmission C1B, where C1 is a constant and
  B is the volume of binding update messages
  decided by the handoffs in wireless domain.

28
Performance Metrics

• Fixed portion of the overhead for single binding
  scheme (Contd)
• Processing overhead
• The processing nodes of the control messages are
  fixed in the Internet portion, lets have the
  number to be C2.
• The conceptual formula for this type of overhead
• Cprocess C2B, where C is the number of
  intermediate routers forwarding the binding
  update messages and B is the volume of binding
  update messages which is affected by the handoffs
  in wireless domain.
• Storage overhead
• No storage overhead is incurred in the Internet
  part.

29
Performance Metrics

• Fixed portion of the overhead for XCAST handoff
  scheme
• Two major observations of Internet portion
• Routes that the control information and data
  traffic flow through in the Internet are not
  changed.
• Volume of data traffic in the Internet is not
  changed since data packets will only be
  duplicated after flowing into the wireless domain.

30
Performance Metrics

• Fixed portion of the overhead for XCAST handoff
  scheme (Contd)
• Based on the observations, we have a constant
  number C3 of all involved nodes in Internet.
• Transmission overhead
• Control messages and headers are not changed in
  Internet.
• Control messages are decided by the handoffs
• Control headers are decided by the throughput and
  ratio of control header
• The cost in the Internet portion can be
  characterized by this conceptual formula C3
  C3V C3H.
• Where C3 is a constant, V is the volume of
  transmission overhead that is affected by
  handoffs in wireless domain, and H is a volume of
  control header overheads which is not decided by
  the handoffs.

31
Performance Metrics

• Fixed portion of the overhead for XCAST handoff
  scheme (Contd)
• Processing overhead
• According to the same factors we listed in the
  previous slide, we have
• number of processing nodes involved by the
  control messages and control headers in the
  Internet is not changed.
• Similar conceptual formula applies for this
  processing overhead C3 C3Proc, where Proc is
  the variable portion affected by the handoffs in
  wireless domain.

32
Performance Metrics

• Fixed portion of the overhead for XCAST handoff
  scheme (Contd)
• Storage overhead
• Since no intermediate status is needed in
  Internet, no storage overhead is incurred in the
  Internet portion.

33
Performance Metrics

• Fixed portion of the overhead for ISM
• Several major observations
• Group membership information does not change with
  the handoffs in the wireless domain.
• Note only for single layer scheme, for
  multi-layer scheme, dynamic group membership
  updates will propagate into the Internet area.
• The Binding Update frequency is not affected by
  the handoffs inside the wireless domain.
• The number of nodes involved in the Inter-Domain
  Multicast Routing protocol is not affected by the
  handoffs.
• Data packets are not duplicated until entering
  the wireless domain.

34
Performance Metrics

• Fixed portion of the overhead for ISM (Contd)
• Based on the above observation, we analyze the
  overhead as below
• Transmission overhead
• Control overhead C in ISM is the membership
  updating and binding update messages sent out by
  the wireless gateway. They can be synchronized
  since they both only indicate the existence of MN
  in the specific domain.
• Data traffic dont have duplicates outside.
• So the conceptual overhead of transmission is
• CISM1 NC LC, where N is the total number
  of multicast routers involved and L is the length
  from wireless gateway to the CN.
• Note the frequency of such control messages is
  lower than frequency of handoffs but N is much
  larger than the number of involved routers in
  XCAST and single binding schemes.

35
Performance Metrics

• Fixed portion of the overhead for ISM (Contd)
• Processing overhead
• Each control information needs to be processed
  and the number of the processing nodes N is not
  changed by the handoffs inside the wireless
  domain.
• Also, each binding update message needs to be
  processed and the number of processing nodes L is
  not changed either.
• So the conceptual formula for the processing
  overhead is the same as the transmission overhead
  formula.

36
Performance Metrics

• Fixed portion of the overhead for ISM (Contd)
• Storage overhead
• Storage is needed in each multicast routers in
  the Internet backbone and the number of multicast
  routers is N.
• For each mobile node, the multicast entry needed
  is a constant 1.
• Note only for single layer scheme, for
  multi-layer scheme, each mobile node may need
  dynamic number of multicast entries in the
  Internet routers.
• So the conceptual formula for all the storage can
  be described as CstorageN.

37
Simulation Environment

• Source traffic at Corresponding Node
• Single rate streaming media
• Encoding rate Rc is dynamically adjusted
  according to the total available bandwidth BW and
  the required redundancy ratio R Rc BW/R.
• Assume a persistent source.
• Always transmit the number of packets according
  to the maximum window size of all the possible
  paths.

38
Simulation Environment

• Source traffic at Corresponding Node (contd)
• Multi-layer multi-rate streaming media
• Assume a dynamic encoding mechanism is employed
  so that at every moment, if currently estimated
  rate for all the path pi is ri, and ri lt rj if i
  lt j, then encoding rate for each layer li
  ri-ri-1, and assume r0 0.
• Assume a persistent source for each video layer.

39
Simulation Environment

• Network configuration
• Wireless infrastructure a network connecting all
  the base stations and routers in the wireless
  area together forms a quasi-tree structure
  network
• Internet wired network connecting networks
  together world widely is modeled by simplified
  channels connecting wireless Gateways directly to
  HA and CNs default router.

40
Network Configuration
Home Agent
Internet
Wireless Gateway
Corresponding Node
Base station
Routers
41
Simulation Environment

• Annotation to the network configuration
• Base station covers a cell cell size is fixed
  and is decided by the radius of the cell rc.
• Distance between any two base stations is the
  same and is d.
• For each base station, different values of
  available bandwidth is assigned to the MN.
• The links between routers have enough capacity to
  carry all the traffic in the simulation, and thus
  congestion may happen only in BS.

d
rc
42
Simulation Environment

• Annotation to the network configuration
• Justification of simplifying the Internet part
• Compared to the limited wireless bandwidth, we
  assume that the available bandwidth in the wired
  part is enough.
• Assume the routes in the Internet part will not
  be affected by the node movement in the wireless
  network.
• We first want to investigate the handoff
  performance within a domain
• If multicast group management is considered, when
  the wireless gateway is chosen as the only
  designated router for the whole domain, the
  multicast routing in Internet part will not be
  affected. (Thus we ignore the overhead of
  inter-domain membership advertisement)

43
Simulation Environment

• Simulating Entities
• One Mobile Node with multiple COA binding
  interfaces
• One Home Agent
• One Corresponding Node
• Intermediate routers in the wireless domain
• One wireless gateway between the wire lined
  Internet and the wireless network
• Channels connecting HA and CN router to wireless
  gateway
• A link connecting CN and the router
• Wireless links with bit error rate of O(10-6)

44
Simulation Environment

• Node mobility model
• The model randomly chooses ?and v(average value
  of Vavg) for a mobile node.
• The mobile node keeps that speed and direction
  for a randomly chosen time t.
• Mobile node keeps moving without any pause time.
• The random speed and direction is generated again.

v
?
45
Simulation Environment

• Available bandwidth distribution
• Available bandwidth on the link from a base
  station to the upper level router is the same for
  all the links of that type.
• Available bandwidth within each cell is
  determined by a random variable with hot spot
  based 2-dimensional Gaussion distribution.
• Available bandwidth on the link between CN and
  its router assume we have enough bandwidth on
  that link.

46
Simulation Environment

• Parameters of the simulation
• Node movement speed
• Vary Vavg in the mobility model.
• Round Trip Time from CN to MN
• Assume each link in wireless area has fixed
  propagation delay.
• Vary the propagation delay on the link from CN to
  the wireless Gateway.

47
Simulation Environment

• Parameters of the simulation (contd)
• Maximum number of COA bindings acquired
  simultaneously.
• By geometrical limitation, the maximum number of
  COA bindings can be set up to 7 (considering the
  hexagon shape of the cell).
• If more COAs than the maximum number are
  available, randomly choose the COAs up to the
  maximum number.

48
Expected Results

• Average blackout time
• XCAST handoff can eliminate blackout time when
  node speed is lower than a certain point.
• After that, node moves so fast that XCAST handoff
  degrades to single binding of COA with the
  blackout time growing up to the round trip time.
• Its possible to apply fast mobile IP handoff
  scheme separately to each binding in XCAST. Thus,
  the performance of XCAST will degrade to the same
  as single binding with fast mobile IP handoff
  scheme.

49
Expected Results

• Frequency of blackout time
• Two single binding schemes have similar blackout
  frequency and it grows as the node mobility
  increases.
• XCAST has really low blackout frequency(nearly
  0) when the node speed is within some threshold
  but the blackout frequency jump to the single
  binding scheme when the node speed is beyond that
  threshold.

Frequency (1/s)
Single binding w/o fast mobile IP support
Single binding w/ fast mobile IP support
1/ RTT
XCAST supported handoff
Node speed (m/s)
Failure point of XCAST
Failure point of fast mobile IP handoff
50
Expected Results

• Throughput
• A. Varying node movement speed.
• Single binding with fast mobile IP handoff
  support always has blackout time and as the node
  speed grows larger, the ratio of effective
  transmission time is smaller.
• Single binding without fast mobile IP support
  also has the same problem but the ratio of
  effective transmission is even worse.
• XCAST handoff can achieve quite high throughput
  when multiple bindings are still maintained. But
  after the point that multiple bindings is not
  possible, it degrades to single binding with
  longer blackout time.
• ISM is discussed later.
• Again, if apply fast mobile IP handoff to each
  COA binding in XCAST, XCAST will degrade to
  single binding with fast mobile IP handoff
  support.

Single binding w/o fast mobile IP support
Single binding w/ fast mobile IP support
XCAST supported handoff
ISM mobility support with single multicast traffic
Ave. throughput (Kbps)
Node Speed (m/s)
51
Expected Results

• Throughput
• B. Varying round trip time

52
Expected Results

• Throughput
• B. Varying round trip time(contd)
• Single binding maintains the typical relationship
  of throughput and round trip time
  .
• XCAST handoff can achieve higher throughput when
  multiple bindings are still maintained. But as
  the round trip time grows up, it finally degrades
  to single binding unicast without fast mobile IP
  handoff support.
• ISM is discussed later.
• Again, if apply fast mobile IP handoff to each
  COA binding in XCAST, XCAST performance will
  degrade to single binding with fast mobile IP
  handoff support.

53
Expected Results

• Throughput
• Conclusion of case A and case B
• Major factor affecting the performance of XCAST
  handoff is the node mobility vs round trip time.
• - If the round trip time is shorter enough for
  the mobile node to acquire an additional binding
  of COA before the node moves out of the previous
  cell, XCAST can give a higher throughput than
  single binding and ISM mobility support scheme.
• - If the round trip time is longer than the
  threshold that the node moves out of the
  transmission range of the first base station
  before it actually registers the new binding
  and start receiving packets through the new
  binding, XCAST handoff will degrade to single
  binding handoff scheme. Thus the performance of
  XCAST handoff will drop below the other two
  schemes.

54
Expected Results

• Uncontrolled traffic
• A. Changing Round Trip Time
• Single binding without fast mobile IP support and
  XCAST handoff scheme will not incur uncontrolled
  traffic because they start transmitting traffic
  to the new link after the sender knows the new
  link.
• Single binding with fast mobile IP support may
  have the shortest connectivity recovery time but
  the uncontrolled traffic it incurs into the new
  link may be the largest because the uncontrolled
  time RTT- Connectivity Recovery Time is the
  largest.
• ISM support also has a short connectivity
  recovery time and has longer uncontrolled time
  than XCAST. If the number of new links is larger
  than 2, ISM may incur more uncontrolled traffic
  than single binding with fast mobile IP support.

Single binding w/o fast mobile IP support
Single binding w/ fast mobile IP support
Uncontrolled traffic (bytes)
XCAST supported handoff
ISM mobility support with single multicast traffic
Round Trip Time (s)
55
Expected Results

• Uncontrolled traffic (Contd)
• B. Changing maximum number of COA bindings
• Because XCAST handoff employs per path congestion
  control, changing the number of COAs will not
  affect the volume of controlled traffic.
• The volume of uncontrolled traffic in ISM will
  increase with the max number of COAs because
  uncontrolled traffic is incurred into more new
  links

Single binding w/o fast mobile IP support
Single binding w/ fast mobile IP support
Uncontrolled traffic (bytes)
XCAST supported handoff
ISM mobility support with single multicast traffic
Max COAs
56
Expected Results

• Loss Rate
• a. Varying node movement speed
• Before the node speed is beyond a certain point,
  XCAST handoff can acquire multiple bindings at
  the same time and the loss rate will be
  drastically low compared to single binding.
• After node speed exceeds a certain point, XCAST
  degrades to single binding without fast mobile IP
  handoff support and loss rate will increase
  quickly.
• If apply fast mobile IP handoff support for each
  COA bindings in XCAST, the performance of XCAST
  will degrade to the same as single binding with
  fast mobile IP handoff support.
• ISM mobility support can have lower loss rate
  because the connectivity recovery time is shorter
  than XCAST and single binding without fast mobile
  IP handoff and it also acquires multiple channels
  during handoffs.

57
Expected Results

• Loss Rate
• Varying the Round Trip Time
• Before the round trip time is beyond a certain
  point, XCAST handoff can acquire multiple
  bindings at the same time and the loss rate will
  be drastically low compared to single binding.
• After round trip time exceeds a certain point,
  XCAST degrades to single binding without fast
  mobile IP handoff support and loss rate will
  increase quickly.
• If apply fast mobile IP handoff support for each
  COA bindings in XCAST, the performance of XCAST
  will degrade to the same as single binding with
  fast mobile IP handoff support.
• ISM mobility support can have lower loss rate
  because of the same reason we listed in the
  previous slide.

58
Expected Results

• Loss Rate
• Loss rate for different layers of streaming
  media
• Lower layers will have more COAs assigned for
  transmission, so they are more robust against
  errors and handoffs.

59
Expected Results

• Transmission overheads control messages
• Varying node movement speed
• Single binding without fast mobile IP handoff
  support scheme only needs to update one binding
  so has lower number of control packets.
• Single binding with fast mobile IP handoff needs
  to send additional binding updates to the
  previous base station. Sp has more control
  packets than single binding without fast mobile
  IP handoff support.
• XCAST handoff needs to send multiple binding
  updates to HA and CN. So the number of control
  packets is larger than single binding schemes.
• ISM need IGMP, MOSPF control messages and MOSPF
  messages are flooded to all the network, so it
  has higher number of control packets.

60
Expected Results

• Transmission overheads control messages
• Varying max. no. of COAs
• Two single binding schemes do not use multiple
  COA bindings. So their control messages doesnt
  increase.
• XCAST handoff scheme will have growing number of
  control messages since it has more binding update
  messages to send.
• ISM also will have more number of control
  messages because more IP addresses need to be
  added to the multicast group. So more IGMP and
  MOSPF messages are sent out.

61
Expected Results

• XCAST control header
• The size of XCAST control header grows linearly
  with the Max number of COAs acquired by the
  mobile node.

Size of XCAST header
Max COAs
62
Expected Results

• Transmission Overhead Data traffic overhead
• Useless traffic
• In single binding with datalink handoff support,
  only unbuffered packets during handoff are
  counted as useless traffic, so the volume of
  useless traffic is a little bit lower than XCAST
  handoff scheme.
• XCAST has a low volume of useless traffic because
  the transmission time of useless traffic is
  limited by the round trip time. Its the same as
  in single binding without datalink support.
• ISM has high volume of traffic because of the
  long leaving delay of the multicast branches.

63
Expected Results

• Transmission Overhead Data traffic overhead
• a. Varying Round Trip Time
• Useless traffic in ISM mobility support is mainly
  decided by the leaving delay of multicast
  branches instead of the round trip time.
• Useless traffic in XCAST handoff scheme is
  proportional to the round trip time and is about
  the same as the single binding without fast
  mobile IP handoff support.
• Useless traffic in single binding with fast
  mobile IP handoff support equals the lost packets
  during handoff and does not change with round
  trip time.

64
Expected Results

• Transmission Overhead Data traffic overhead
• a. Varying node mobility
• Useless traffic in ISM mobility support is mainly
  decided by the leaving delay of multicast group.
  So it remains the same.
• In single binding schemes, useless traffic starts
  growing even the node mobility is slow. Thats
  because they only receive packets through one
  binding at any time.
• When node mobility grows large, single binding
  with fast mobile IP handoff support will have
  smaller size of useless traffic because it takes
  the shortest time to recover the connectivity.
• XCAST will not have useless traffic when node
  mobility is lower than a threshold. Beyond that,
  it performs similar to single binding scheme.

65
Expected Results
Single binding w/o datalink support

• Processing overhead
• Single binding with datalink handoff support will
  have higher computation overhead because it needs
  to transfer the buffered packets.
• Single binding without datalink support will have
  the lowest computation overhead because it needs
  nothing except for the binding updates.
• ISM also has higher processing overhead because
  the computation of multicast tree and processing
  of membership updates.
• XCAST handoff has lower processing overhead
  because
• all processing is at end system and HA
• major operation is binding updates

Single binding w/ datalink support
XCAST supported handoff
ISM mobility support with single multicast traffic
Computation overhead
Node Mobility (m/s)
66
Expected Results

• Overhead
• Conclusion
• XCAST handoff has the moderate number of control
  packets and useless traffic among the three
  handoff schemes.
• XCAST handoff needs the least processing overhead
  among the three.
• Max number of COAs also has impact on the XCAST
  scheme because it affects the number of binding
  update messages and the size of the XCAST control
  header.
• Parameters like node mobility has similar impact
  to the control overhead of all three schemes, but
  round trip time only has impact on XCAST handoff
  scheme.

67
XCAST vs. ISM Mobility Support

• Requirement of the fast handoff scheme
• smoothly handle fast handoff with quick recovery
  of connectivity and transmission rate
• reduce data loss during handoffs
• keep high throughput during handoffs
• make rate sensitive application sustainable in
  fast handoff situation

68
XCAST vs. ISM Mobility Support

• Fast connectivity recovery
• ISMhave an intermediate node on the previous
  path relay packets to the new path without
  informing CN or HA
• XCAST keep multiple bindings simultaneously so
  that moving away from one cell does not cause
  complete connectivity loss

69
XCAST vs. ISM Mobility Support

• Reduce data loss during handoffs
• ISM keep multiple paths during handoffs and have
  duplicate copies to be delivered
• XCAST

70
XCAST vs. ISM Mobility Support

• High throughput during handoffs
• ISM
• XCAST

71
XCAST vs. ISM Mobility Support

• Friendliness to rate sensative flows
• ISM
• XCAST

72
XCAST vs. ISM Mobility Support

• Scalability with respect to the number of mobile
  nodes
• Intermediate router support/overhead

73
XCAST vs. ISM Mobility Support

• Additional objectives achieved by the proposed
  XCAST handoff scheme
• Application specific utilization of bandwidth
  available over multiple paths
• Multi-layer streaming media
• Single layer streaming media

74
XCAST vs. ISM Mobility Support

• 4.2.1 Argument for using XCAST instead of ISM in
  mobility support(Contd)
• Using multicast only to support fast handoff can
  achieve the goal to fast recover the
  connectivity. The multicast group supporting
  mobility should have the following features
• Membership of the multicast group is highly
  dynamic because each member address is a
  dynamically bound COA.
• Number of members is limited because the number
  of COAs the mobile device can acquire is limited.
• The technique should be scalable in terms of
  number of group addresses because each mobile
  node will need at least one multicast group
  address to support its mobility.

75
Performance of XCAST handoff scheme and Impacts
of Different Parameters

• throughput in a time scale

XCAST handoff throughput
Throughput (Kbps)
XCAST handoff convergence period to fair
bandwidth share
Single binding handoff scheme
ISM with single rate streaming
Time (s)
76
Performance of XCAST handoff scheme and Impacts
of Different Parameters

• throughput in a time scale(Contd)
• XCAST handoff scheme has the most smooth
  transmission rate in handoff among the three
  schemes.
• XCAST handoff scheme also acquires highest
  throughput among all the three schemes.

77
Performance of XCAST handoff scheme and Impacts
of Different Parameters

• Impact of Max COAs and node mobility to the
  throughput
• Conclusion hopefully, by adding more COA
  bindings, the possibility for the mobile node to
  acquire higher bandwidth will increase. So does
  the throughput. But we may only see the
  performance increase by adding a few number of
  COAs.

78
Performance of XCAST handoff scheme and Impacts
of Different Parameters

• Impact of Max COAs and node mobility to the loss
  rate
• Conclusion by adding number of COA bindings, we
  add the redundancy of the transmission. So the
  loss rate will decreases. But again, we probably
  can only see the benefits by adding a few COAs.

79
Performance of XCAST handoff scheme and Impacts
of Different Parameters

• Impact of Node mobility and Round Trip Time to
  the throughput
• Conclusion XCAST handoff has quite good and
  stable performance before the node mobility
  exceeds a threshold.

80
Performance of XCAST handoff scheme and Impacts
of Different Parameters

• Impact of RTT and node mobility to the blackout
  time

Blackout Time
Failure points
RTT4
RTT3
RTT2
RTT1
Node Mobility (m/s)
81
Performance of XCAST handoff scheme and Impacts
of Different Parameters

• Conclusion about the impact of RTT and node
  mobility
• Round trip time plays an important role in the
  performance of XCAST handoff scheme. It will
  affects the throughput, blackout time,
  convergence time, and control overhead of XCAST
  scheme.
• So, find the limitation of the round trip time
  under some type of node mobility is important in
  measuring the performance of XCAST handoff scheme.