Transcript Receiver-driven Layered Multicast

Receiver-driven Layered
Multicast
Paper by- Steven McCanne, Van Jacobson
and Martin Vetterli – ACM SIGCOMM 1996
Presented By – Manoj Sivakumar
Overview






Introduction
Approaches to Rate-Adaptive
Multimedia
Issues and challenges
RLM - Details
Performance Evaluation
Conclusions
Introduction

Consider a typical streaming
Application
Source

128 Kb/s
Internet
X Kb/s
Receiver
What rate should the source send data
at ?
Approaches to Rate-Adaptive
Multimedia

Rate Adaptation at Source – based on
available network capacity


Works well for a Unicast environment
How about multicast ?
Receiver 1
X1 Kb/s
source
128 Kb/s
X2 Kb/s
X3 Kb/s
Receiver 2
Receiver 3
Example of Heterogeneity
Issues and Challenges




Optimal link utilization
Best possible service to all receivers
Ability to cope with Congestion in the
network
All this should be done with just best
effort service on the internet
Layered Approach


Rather than sending a single encoded
video signal the source sends several
layers of encoded signal – each layer
incrementally refining the quality of the
signal
Intermediate Routers drop higher
layers when congestion occurs
Layered Approach


Each layer is sent to one multicast
group
If a receiver wants higher quality –
subscribes to all higher level layer
multicast groups
Issue in Layered Approach


No framework for explicit signaling
between the receivers and routers
A mechanism to adapt to both static
heterogeneity and dynamic variations
in network capacity is not present

Solution - RLM
RLM – Network Model


Works with IP Multicast
Assume




Best effort (packets may be out of order, lost or
arbitrarily delayed)
Multicast (traffic flows only along links with
downstream recipients)
Group oriented communication (senders do not
know of receivers and receivers can come and
go)
Receivers may specify different senders
RLM - Video Streams




One channel per layer
Layers are additive
Adding more channels gives better
quality
Adding more channels requires more
bandwidth
RLM Sessions


Each session composed of layers, with
one layer per group
Layers can be separate (i.e. each layer
is higher quality) or additive (add all to
get maximum quality)

Additive is more efficient
Router Mechanisms

Dropping of packets

Drop less preferential packets first
RLM - Protocol

Abstraction


on congestion, drop a layer
on spare capacity, add a layer
RLM – Adding and Dropping
layers



Drop layer when packet loss
Add does not have counter-part signal
Need to try adding at well-chosen
times

Called join experiment
RLM – Adding and Dropping
layers

If join experiment fails


If join experiment succeeds


Drop layer, since causing congestion
One step closer to operating level
But join experiments can cause
congestion

Only want to try when might succeed
RLM – Join Experiments

Get lowest layer and start timer for
next probe



Initially timer small
If higher level fails then increase timer
duration else proceed to next layer and
start time for the layer above it
Repeat until optimum
RLM Join Experiment

How to know is join experiment
succeeded

Detection time
Detection Time




Hard to estimate
Can only be done experimentally
Initially start with a large value
Progressively update the detection
time based on actual values
RLM - Issues with Joins

Is this Scalable



What if each node does join experiments
and the same time for different layers
Wrong info to node that requests lower
layer if the other node had requested
higher layer
Solution – Shared Learning
RLM – Shared Learning

Each node broadcasts its intent to the
group



Adv’s – other nodes can learn from the
result of this node’s experiment
Reduction in simultaneous experiments
Is this still foolproof ??
RLM - Evaluation



Simulations performed in NS
Video modeled as CBR
Parameters







Bandwidth: 1.5 Mbps
Layers: 6, each 32 x 2m kbps (m = 0 … 5)
Queue management :Drop Tail
Queue Size (20 packets)
Packet size (1 Kbytes)
Latency (varies)
Topology (next slide)
RLM - Evaluation

Topologies




1 – explore latency
2 – explore
scalability
3 – heterogeneous
with two sets
4 – large number of
independent
sessions
RLM – Performance Metrics

Worse-case lost rate over varying time intervals



Throughput as percent of available



But will always be 100% eventually
So, look at time to reach optimal
Note, neither alone is ok



Short-term: how bad transient congestion is
Long-term: how often congestion occurs
Could have low loss, low throughput
High loss, high throughput
Need to look at both
RLM – Performance Results

Latency Results
RLM – Performance Results

Latency Results
RLM – Performance Results

Session Size
RLM – Performance Results

Convergence rate
RLM – Performance Results

Bandwidth Heterogeneity
Conclusions

Possible Pitfalls




Shared Learning assumes only multicast
traffic
Is this valid ??
Is congestion produced by Multicast
traffic alone
Simulation does not other traffic
requests!!
Conclusions


Overall – a nice architecture and
mechanism to regulate traffic and have
the best utilization
But still needs refinement
References

S. McCanne, V. Jacobson, and M.
Vetterli, "Receiver-driven layered
multicast," in Proc. SIGCOMM'96,
ACM, Stanford, CA, Aug. 1996, pp.
117--130.