Preemptive Strategies to Improve Routing Performance of Native and Overlay Layers

1
Preemptive Strategies to Improve Routing
Performance ofNative and Overlay Layers

• Srinivasan Seetharaman - College of Computing,
  Georgia Tech
• Volker Hilt - Multimedia Networking, Bell Labs
• Markus Hofmann - Multimedia Networking, Bell
  Labs
• Mostafa Ammar - College of Computing, Georgia
  Tech

2
Multi-Layer Interaction

• Service overlay networks offer enhanced services
  by forming a virtual network of specialized nodes
• They deploy independent routing schemes that
• are oblivious to underlying native network
• achieve a specific selfish objective
• Two main problems
• Mismatch of routing objectives
• Misdirection of traffic matrix estimation

3
Repeated Game Model

• Player1 Overlay Routing (OR)
• Latency-optimized paths between nodes
• Reacts to changes in link latency by probing
  periodically, without concern for bandwidth
• Player2 Traffic Engineering (TE)
• MPLS-based scheme that solves a linear program
  (using GNU LP kit) to obtain optimal multi-paths
  using traffic matrix as input
• Minimize Max util Maxa?E ( Xa/Ca )

4
Repeated Game model (contd.)
Overlay Routing
Overlay routes
Overlay Link Latencies
Overlay layer traffic
?
Native link delays
?
Traffic on each overlay link
Traffic Engineering
Native routes
Background traffic
?
TrafficMatrix
5
Illustration of OR vs TE
C
14ms

Shortest latency routes

A
4ms


4ms
5ms
B

10ms
D

23ms
OVERLAY
NATIVE
2
F
E
4
10ms
2ms
C
4
3
3
3
4
2ms
2ms
Minimize(Max util)
2ms
Numbers on each link represent the avail-bw
3ms
2ms
B
5
A
H
G
4ms
2
3
2
3
3ms
6ms
2ms
3ms
2
2
I
J
D
10ms
10ms
Initial State
6
Illustration of OR vs TE (contd.)
C
14ms

Multihop paths A ? B ? C A ? B ? D

A
6ms


4ms
5ms
B

10ms
D

23ms
OVERLAY
NATIVE
2
F
E
4
10ms
2ms
C
4
0
0
2
2
2ms
2ms
2ms
3ms
2ms
1
B
A
H
G
4ms
1
2
2
2
3ms
6ms
2ms
3ms
2
2
I
J
D
10ms
10ms
Overlay traffic introduced
Avail-bw changed
7
Illustration of OR vs TE (contd.)
C
14ms

Multihop paths A ? B ? C A ? B ? D

A
4ms


5ms
5ms
B

10ms
D

23ms
OVERLAY
NATIVE
2
F
E
2
10ms
2ms
C
2
1
1
2
4
2ms
2ms
2ms
3ms
2ms
B
3
SPLIT
A
H
G
4ms
1
1
1
2
3ms
6ms
2ms
3ms
2
2
I
J
D
10ms
10ms
Latencychanged
After TE reacts
8
Illustration of OR vs TE (contd.)
C
14ms

Multihop paths A ? B ? C A ? B ? C ? D B ? C ? D

A
4ms


5ms
5ms
B

10ms
D

23ms
OVERLAY
NATIVE
2
F
E
0
10ms
2ms
C
0
1
1
0
4
2ms
2ms
2ms
3ms
2ms
B
5
SPLIT
A
H
G
4ms
1
3
1
0
3ms
6ms
2ms
3ms
2
2
I
J
D
10ms
10ms
After Overlay routing reacts
Avail-bw changed
9
Simulation Setup

• Random GT-ITM generated topologies
• Overlay network 5 nodes (or) 8 nodes
• Native network 20 nodes
• Total combinations simulated 50 topologies
• Random static traffic assignment
• Fix total load by tuning the avg link util
• Determine amount of overlay traffic
• Remaining Background traffic
• Every TE event is followed by 3 OR events

10
Simulation Results

• TEobjective
• Overlayobjective
• Overallstability

?Round?
11
Past research

• Qiu-Sigcomm03 conducted a simulation study of
  scenarios where there is a conflict of objectives
• Liu-Infocom05 analyzed the interaction between
  OR and TE to show existence of Nash equilibrium
• General conclusion
• The system suffers from prolonged route
  oscillations and sub-optimal routing costs

12
Our goal

• .. is to propose strategies that
• obtain the best possible performance for a
  particular layer
• while steering the system towards a stable state.

13
Resolving Conflict Basic Idea

• Designate leader / follower
• Leader will act after predicting or counteracting
  the subsequentreaction of the follower
• Similar to the Stackelberg approach

14
Resolving Conflict - Obstacles

• Incomplete information
• Unavailable relation between the objectives
• NP-hard prediction

15
Resolving Conflict - Simplification

• Assume Each layer has a general notion of the
  other layers selfish objective
• Operate leader such that
• Follower has no desire to change ? Friendly
• Follower has no alternative to pick ? Hostile
• Constitutes a preemptive action
• Use history to learn desired action gradually.

16
Overlay Strategy - Friendly

• Native layer only sees a set of src-dest demands
• Improve latency of overlay routes, while
  retaining the same load pressure on the native
  network!
• Load-constrained LP

C
1
E
B
D
1
A
17
Overlay Strategy Friendly (contd.)
Acceptable to both OR and TE
Stable within a few rounds
18
Overlay Strategy - Hostile

• Push TE to such an extent that it does not
  reroute the overlay links after overlay routing
• Send dummy traffic in an effort to render TE
  ineffective
• Dummy traffic injection

C
1
E
Unused overlay link AB
B
D
1
A
19
Overlay Strategy - Hostile (contd.)
TE cant improve further
Acceptable only to OR
20
Native Strategy - Friendly

• TE pays no attention to the length of the route!
• TE should balance load, while ensuring that the
  path length is almost the same!
• Hopcount-constrained LP

C
1
E
B
D
1
A
21
Native Strategy - Friendly (contd.)
Acceptable to both OR and TE
Takes a bit longer to converge
22
Native Strategy - Hostile

• Dissuade overlay routing from using certain
  multihop paths
• Increase latency of native links that are heavily
  loaded, without any knowledge of overlay networks
• Load-based latency tuning

Overusednative link
C
1
E
1
B
D
1
A
23
Native Strategy - Hostile (contd.)
Disrupted overlay routing
Takes a bit longer to converge
24
Preemptive Strategies Summary

• We proposed four strategies that improve
  performance for one layer and achieve a stable
  operating point
• Inflation factor
• Steady state obj value with strategy
• Best obj value achieved

Inflation
25
Preemptive Strategies Summary (contd.)

• Each strategy achieves best performance for the
  target layer
• within a few rounds
• with no interface between the two layers
• with all information inferred through simple
  measurements
• If both layers deploy preemptive strategies, the
  performance of each layer depends on the other
  layers strategy.