Multicasting

Download Report

Transcript Multicasting

Host Multicast: A Framework for
Delivering Multicast to End Users
Jonghak Kim
Hyungik Oh
Huo Da
1
June, 2002
INFOCOM
Motivation


Fast increasing need for scalable and efficient
group communication
Slow deployment of IP Multicast
–

Emerging End-host based Multicast
–
–
2
Deployment imposes dependency on routers
Member hosts duplicate and forward packets
Easy to deploy, but less efficient
Host Multicast



A hybrid approach
Goal: Ubiquitous Multicast
Design Requirements
–
Deployable on the current Internet


–
Compatible with IP Multicast to the furthest extent


–
Use IP Multicast where available
Provide incentive to future deploy
Scalability and Efficiency

3
No support is required from OS, routers and servers
Enable multicast applications
Multicast group should not be a limiting factor
Architecture
Rendezvous Point (HMRP)
RP
Designated Member (DM)
DM
Host
Host
Host
DM
Host
Unicast Tunnel
DM
Host
Host
Normal Member
IP Multicast Island
4
Host
Group
Management
Protocol
(HGMP)
for
intra-island
A network
Each
To
bootstrap
member
Multicast
of
new
any
runs
Tree
size
members
our
Protocol
that
daemon
supports
(HMTP)
program
IPtoMulticast,
atbuild
user-space
inter-island
e.g. single
tunnels
host,
management
Ethernet, campus network etc.
Host Multicast Tree Protocol (HMTP)


Build a bi-directional shared-tree connecting all
islands
The tree should be congruent to physical
network topology to be efficient
–

The tree should be robust
–
5
use member-to-member round-trip time as distance
metric in current design
be able to handle node failure, dynamic join/leave
etc.
Join Group
root
HMRP always knows
the root of the tree.
B
A newcomer does a
depth-first search of
the tree to find a close
member as its parent.
A
C
D
F
E
6
Clustering nearby
members makes the
tree congruent to
physical network
topology to the first
order.
H
G
Where is my group?
Root of your group is A
RP
Tree Maintenance


7
Each member keeps
its children list and
root path up to date
by exchanging
REFRESH and
PATH messages
with neighbors.
Root sends
REFRESH message
to HMRP.
B
A
C
D
F
E
H
G
RP
Member Leave and Partition
Recovery


Parent deletes the
leaving node from
children list.
Direct Children repair
the tree by running
join procedure in
the reverse order.

8
If root is leaving, the
first node contacting
HMRP is assigned as
new root.
B
A
C
D
F
E
H
G
RP
Tree Improvement

Periodically re-run the join procedure
–
–
–
9
To accommodate changes in network conditions
and group membership
Start from a randomly picked node in the root path.
Less frequent than REFRESH and PATH
messages.
Loop Detection and Resolution

Loop is possible:
–

10
A
C
D
One’s root path contains itself
Resolution:
–

B
Detection:
–

Multiple conflicting joins
happen at the same time.
Leave the current parent and
re-join the tree from the root.
Loop is rare.
E
F
G
~ The End ~
11