Download ppt

15-744: Computer Networking
L-16 Multicast Reliability and
Congestion Control
Multicast Issues
• Reliable multicast
• Multicast congestion control
• Assigned reading
• [F+97] A Reliable Multicast Framework for
Light-Weight Sessions and Application Level
Framing
© Srinivasan Seshan, 2002
L -16; 11-5-02
2
Overview
• Scalable Reliable Multicast
• Congestion Control
© Srinivasan Seshan, 2002
L -16; 11-5-02
3
SRM
• Originally designed for wb
• Receiver-reliable
• NACK-based
• Every member may multicast NACK or
retransmission
© Srinivasan Seshan, 2002
L -16; 11-5-02
4
SRM Request Suppression
Packet 1 is lost; R1 requests
resend to Source and Receivers
Resend request
Packet 1 is resent; R2 and R3 no
longer have to request a resend
Resent packet
S
X1
2
1
R1
R1
R2
R3
© Srinivasan Seshan, 2002
S
Delay varies
by distance
R2
R3
L -16; 11-5-02
X
X
5
Deterministic Suppression
3d
d
data
Time
2d
d
d
d
nack
3d
repair
= Sender
= Repairer
d
4d
= Requestor
d
Delay = C1dS,R
© Srinivasan Seshan, 2002
L -16; 11-5-02
6
SRM Star Topology
Packet 1 is lost; All Receivers
request resends
Resend request
Resent packet
S
X1
R2
Packet 1 is resent to all Receivers
S
2
R3
1
R4
R2
R3
R4
Delay is same length
© Srinivasan Seshan, 2002
L -16; 11-5-02
7
SRM: Stochastic Suppression
0
d
d
Time
data
1
repair
session msg
d
2
NACK
d
3
2d
= Sender
Delay = U[0,D2] dS,R
= Repairer
= Requestor
© Srinivasan Seshan, 2002
L -16; 11-5-02
8
SRM (Summary)
• NACK/Retransmission suppression
• Delay before sending
• Delay based on RTT estimation
• Deterministic + Stochastic components
• Periodic session messages
• Full reliability
• Estimation of distance matrix among members
© Srinivasan Seshan, 2002
L -16; 11-5-02
9
What’s Missing?
• Losses at link (A,C)
causes retransmission
to the whole group
• Only retransmit to
those members who
lost the packet
• [Only request from the
nearest responder]
S
0
0
B
A
0.99
C
0
0
D
Sender
© Srinivasan Seshan, 2002
L -16; 11-5-02
E
0
F
Receiver
10
Local Recovery
• Different techniques in various systems
• Application-level hierarchy
• Fixed v.s. dynamic
• TTL scoped multicast
• Router supported
© Srinivasan Seshan, 2002
L -16; 11-5-02
11
Overview
• Scalable Reliable Multicast
• Congestion Control
© Srinivasan Seshan, 2002
L -16; 11-5-02
12
Multicast Congestion Control
• What if receivers
have very
different
bandwidths?
• Send at max?
• Send at min?
• Send at avg?
100Mb/s
100Mb/s
R
S
1Mb/s
???Mb/s
R
1Mb/s
56Kb/s
© Srinivasan Seshan, 2002
L -16; 11-5-02
R
R
13
Video Adaptation: RLM
•
•
•
•
•
•
Receiver-driven Layered Multicast
Layered video encoding
Each layer uses its own mcast group
On spare capacity, receivers add a layer
On congestion, receivers drop a layer
Join experiments used for shared learning
© Srinivasan Seshan, 2002
L -16; 11-5-02
14
Layered Media Streams
R2
R1
R1 joins layer 1,
joins layer 2
joins layer 3
10Mbps
10Mbps
S
R
10Mbps
R2 join layer 1,
join layer 2
fails at layer 3
512Kbps
R
128Kbps
R3
© Srinivasan Seshan, 2002
L -16; 11-5-02
R3 joins layer 1,
fails at layer 2
15
Drop Policies for Layered Multicast
• Priority
• Packets for low bandwidth layers are kept, drop
queued packets for higher layers
• Requires router support
• Uniform (e.g., drop tail, RED)
• Packets arriving at congested router are
dropped regardless of their layer
• Which is better?
• Intuition vs. reality!
© Srinivasan Seshan, 2002
L -16; 11-5-02
16
RLM Intuition
Uniform vs. Priority Dropping
70
Performance
60
50
Uniform
Priority
40
30
20
10
0
© Srinivasan Seshan, 2002
Offered load
L -16; 11-5-02
17
RLM Intuition
• Uniform
•
•
•
•
Better incentives to well-behaved users
If oversend, performance rapidly degrades
Clearer congestion signal
Allows shared learning
• Priority
• Can waste upstream resources
• Hard to deploy
• RLM approaches optimal operating point
• Uniform is already deployed
• Assume no special router support
© Srinivasan Seshan, 2002
L -16; 11-5-02
18
RLM Join Experiment
• Receivers periodically try subscribing to
higher layer
• If enough capacity, no congestion, no drops
 Keep layer (& try next layer)
• If not enough capacity, congestion, drops
 Drop layer (& increase time to next retry)
• What about impact on other receivers?
© Srinivasan Seshan, 2002
L -16; 11-5-02
19
Join Experiments
Layer
4
3
2
1
Time
© Srinivasan Seshan, 2002
L -16; 11-5-02
20
RLM Scalability?
• What happens with more receivers?
• Increased frequency of experiments?
• More likely to conflict (false signals)
• Network spends more time congested
• Reduce # of experiments per host?
• Takes longer to converge
• Receivers coordinate to improve behavior
© Srinivasan Seshan, 2002
L -16; 11-5-02
21
Next Lecture: Security
•
•
•
•
Denial of service
IPSec
Firewalls
Assigned reading
• [SWKA00] Practical Network Support for IP
Traceback
• [B89] Security Problems in the TCP/IP Protocol
Suite
© Srinivasan Seshan, 2002
L -16; 11-5-02
22