ALMI An Application level multicast infrastructure

1
ALMIAn Application level multicast
infrastructure
2
General Problems in Multicast

• Traditional IP Multicast requires control support
  from network devices (e.g. IP routers), in
  particular membership management (IGMP) and
  multicast routing protocols.
• IP routers maintains separate routing state for
  each multicast group. This model is relatively
  less scalable with respect to the number of
  concurrently multicast group.

3
General Problems in Multicast (2)

• Widespread deployment of IGMP and routing
  protocols requires substantial infrastructure
  modifications and complex modifications to IP
  Routers Software.
• Some issues of IP Multicast (e.g. end-to-end
  reliability, flow and congestion control
  schemes), offer significant challenges but no
  clear solutions have emerged thus far.

4
Particular Problem

• There are large number of applications that
  contain a small number of group members.E.g.
  video conferencing, multi-party games, private
  chat rooms, etc.
• These groups are often created and destroyed
  relatively dynamically.
• The number of such groups that active in the same
  time can be fairly large.

5
Particular Problem (2)

• For a large number of such small and sparse
  groups, the benefits of IP Multicast in terms of
  bandwidth efficiency and scalability are quiet
  often outweighed by the control complexity (group
  set up and maintenance).
• Due to the increasing number of such applications
  we need multicast protocol which can be supported
  without relying on the IP multicast
  infrastructure.

6
Solution

• We need a solution for
• Multi-sender multicast communication
• Which scales for a large number of communication
  groups with small number of members
• Does not depend on multicast support in the
  routers.
• Then, Application Level Multicast Infrastructure
  (ALMI) addresses to these concerns

7
Overview ALMI

• ALMI offers- accelerated deployment-
  simplified configuration- better access control
  at the cost of additional (even though small)
  traffic load in the network
• ALMI is implemented in the user space, then its
  more flexible in customizing some aspects, e.g.
  data transcoding, error recovery, flow control,
  etc.

8
Overview ALMI (2)

• Participants of a multicast session are connected
  via a virtual multicast tree.
• The tree consists of unicast connections between
  end hosts.
• The tree is formed as a Minimum Spanning Tree
  (MST).
• The cost of each link is an application specific
  metric. (There is no such strict metric)

9
ALMI Architecture

• ALMI session consists of1. session
  controller2. (multiple) session members
• Session Controller and session members are
  program instances.

10
Session Controller

• Session Controller is possibly located at- a
  place that easily accessible by all members.-
  co-located with one of the session members-
  special purpose server or a multicast proxy
  installed within corporate- an ISP network

11
Session Controller (2)

• Session controller handles- member
  registration- maintains the multicast tree
• Controller performs these 2 functions- ensures
  connectivity of the multicast tree when members
  join and/or leave the session and when network
  or host failures occur.

12
Session Controller (3)

• - ensures the efficiency of the multicast tree
  by periodically calculating a minimum spanning
  tree based on the measurement updates received
  from all members.To collect measurements, the
  controller essentially instruct each member to
  monitor a set of other members.

13
Session Members

• Session members are organized into a multicast
  tree
• A link in the multicast tree represents a unicast
  connection between two members
• Session data is disseminated along this multicast
  tree, while control messages are unicast between
  each member and the controller.

14
Session Members (2)

• Multicast tree ? shared tree amongst member with
  bidirectional links.
• Two members connected by link, receive a
  designation of parent and child.
• This parent-child relation does not indicate
  direction of data flow

15
Session Members (3)

• A session member forwards data to designated
  adjacent neighbors.
• A session member also monitors the performance of
  unicast paths to and from a subset of other
  session members.
• This is achieved by periodically sending probes
  to these members and measuring an application
  level performance metric.

16
Session Members (4)

• The measurements are then reported to the
  controller and serve as a the costs used to
  calculate a Minimum Spanning Tree.

17
Fault Tolerance

• Centralized controller architecture would
  constitute a single point of failure. Two points
  should be made in this respect
• First, with multiple back-up controllers, which
  operates in stand-by mode.
• The address of these back-up controllers must be
  known by all members.
• Back-up controllers periodically receive state
  from the primary controller.

18
Fault Tolerance (2)

• Second, even in the case that no control, the
  operation is possible.
• The existing ALMI tree, and hence data path, will
  remain unaffected and will continue the operation
  until a membership change or a critical failure
  occurs.

19
Control Plane Operation

• ALMI relies on a control protocol for
  communication between session controller and
  session members.
• Control protocol handles tasks related to
  membership management, performance monitoring,
  and routing.
• ALMI uses a common packet format to carry both
  data and control packets.

20
Control Plane Operation (2)
ALMI PACKET HEADER FORMAT
21
Control Plane Operation (3)

• Session ID and Source ID are generated by
  controller and guaranteed to be collision free.
• Flag field defines various type of operationsa.
  Registration Messages between host to
  controllerb. Connection Request between parent
  and child.C. Performance Monitoring reported
  from members to session controller

22
Control Plane Operation (4)

• c. Distribution Tree Messages, generated by
  controller, used to inform members of their
  peering points.d. Neighboring Monitoring Update
  Message, sent by controller to members to
  inform them a new list of neighbors they need
  to monitor.e. Departure messages, sent from
  group members to the controller and their
  current parent and children.

23
Control Plane Operation (5)

• Tree Incarnation field is to prevent loops and
  partitions in the multicast tree.
• Tree Updates messages are disseminated
  independently to all members and there is
  possibility that some control messages might be
  lost.
• On the other hand, members act on update messages
  in varying delay.

24
Control Plane Operation (6)

• In order to prevent loop and partition,
  controller assigns a monotonically increasing
  version number to each newly generated multicast
  tree.
• Source generates packets including its latest
  tree incarnation in the packet header.
• Each ALMI member maintain a small cache of recent
  multicast incarnations.
• When members receive packets with tree version
  not contained in the cache, packets will be
  discarded.

25
Control Plane Operation (7)

• If a member receives a data packet with a newer
  tree version, it detects that its information is
  not up to date and therefore re-registers itself
  to the controller to receive new tree information.

26
Member Operation

• Session member can allocate the session
  controllers address and port number through
  online or offline scheme (e.g. URL, directory
  service, email message, etc)
• Host registers by sending JOIN message to session
  controller. Then, session controller will give
  member ID, and its parents ID and address.

27
Member Operation (2)

• The new member sends GRAFT message to its parent,
  asking for data port which it sends and receives
  data
• Data transfer between two adjacent members can be
  reliable or unreliable by deploying TCP or UDP.

28
Member Operation (3)

• If child detects failure of the connection to its
  parent, it consider itself disconnected from the
  graph and send REJOIN message to the controller.
• If parent detects failure of the connection to
  its child, it simply closes the connection

29
Member Operation (4)

• If a member detects through the monitoring
  process that its parent is not responding or
  receives a LEAVE message from the parent, it then
  issues REJOIN message to the controller,
  repeating the steps as when joining an ALMI group.