Overlay Network Creation and Maintenance with Selfish Users

1
Overlay Network Creation and Maintenance with
Selfish Users

• Georgios Smaragdakis

Dissertation committee members Azer Bestavros,
Nikolaos
Laoutaris, John Byers
2
Overlays Neighbor Selection
Overlay node
Overlay links
Internet
Overlay applications overlay routing, p2p file
sharing, content distribution..
Transit ISP
Transit ISP
Focus on service quality!
Access ISP
Access ISP
Access ISP
3
Challenges
v4
v1
v7
v5
v2
v6
v8

• What is the performance gain that
• can be achieved by a selfish node?

p1v2v3v4v5v6v7v8v9
p8v1v2v3v4v5v6v7v9
v3

• What is the impact of selfish neighbor
• selection to overlay network performance?
• What are the implications of selfish neighbor
  selection to system design?

p3v1v2v4v5v6v7v8v9
v9
Selfish node
p9v1v2v3v4v5v6v7v8
4
Outline
Implications to Overlay Routing
Selfish Neighbor Selection
Implications to File Sharing
Implications to Service Provisioning
5
Implications to Overlay Routing
Selfish Neighbor Selection
Implications to File Sharing
Implications to Service Provisioning
6
Selfish Neighbor Selection (SNS)

• Constraints that need to be addressed in a
  realistic model for overlay networks
• Bounded degree
• Preference vectors
• Realistic network distance
• Link directionality
• Fundamentally different from other models that
  have been proposed for other networks.
• Fabrikant et al.,PODC03 Chun et al.,
  Infocom04

7
Optimal Neighbor Selection

• vi choose k neighbors, s.t.

min
over all si?Si
vi
G-i( V-i , S-i )
Set of residual nodes
Set of residual wiring
vis residual network
8
SNS Facility Location

• Uniform link weights, and uniform preference
• ? k-median on asymmetric distances

9
k-median

• k-median
• Find a subset I of F and a function sC?I
• to min ( Si,j sjcij ) such that I k

F set of facilities
C set of clients, cij cost connecting client
j?facility I sj demand of node j
10
Uncapacitated Facility Location

• Uncapacitated Facility Location (UFL)
• Find a subset I of F and a function sC?I
• to min ( Si fi Si,j sjcij )

F set of facilities fi cost to
open facility
C set of clients, cij cost connecting
client j?facility I sj demand of node j
11
SNS Facility Location

• Uniform link weights, and uniform preference
• ? k-median on asymmetric distances

Since the wiring cost is the same

• Non-uniform link weights, and uniform
• preference
• ? ILP formulation

vi
min
12
Local Search (LS)

• vi choose k neighbors

min
w
over all si?Si
u
vi
Arya et al,STOC01
G-i( V-i , S-i )
Set of residual nodes
Set of residual wiring
vis residual network
13
SNS the Game

• Game ltV,si,Cigt
• V set of n players (nodes)
• si strategies available to vi (wirings),
• choose k out of n to connect
• Ci set of costs for vi
• min
• Best response of a node nodes optimal wiring
• Outcome S, the global wiring
• A stable wiring is a pure Nash equilibium
• Using iterative best response
• Fundamentally different from selfish routing

14
SNS Equilibria
Uniform Preference
Skewness of preference

• n15 k2
• k3
• k8
• k11

In-degrees are highly skewed even under
uniform preference ! ? Quality-based
preferential attachment
k (Link density)
15
SNS Efficiency

• Performance of ILP LS is close to Utopian!
• Theoretical results showed in the worst case the
  cosial cost can be bad
• Laoutaris, Poplawsi, Rajaraman,
  Sundaram, Teng,PODC08

Skewness of preference
Skewness of preference
Link density
Link density
16
SNS Trace-Driven Evaluation

• How we assign the distance
• Synthetically using BRITE
• Empirically from PlanetLab
• Empirically from AS-level maps Routeviews
• Neighbor Selection Strategies
• k-Random heuristic
• k-Closest heuristic
• k-Regular heuristic
• k-Best Response
• Control parameter
• Bound on out-degree k (link density)

17
Connecting on a k-Random graph
AS-Level (n50)
PlanetLab (n50)
BRITE (n50)
0 2 3 5 11 22
0 2 3 5 11 22
0 2 3 5 11 22
k
k
k
If your neighbors are naïve, it pays to be
selfish!
18
Connecting on a k-Closest graph
AS-Level (n50)
PlanetLab (n50)
BRITE (n50)
0 2 3 5 11 22
0 2 3 5 11 22
0 2 3 5 11 22
k
k
k

• Greed is not good

If your neighbors are greedy, it pays to be
selfish!
19
Connecting on a k-Regular graph
AS-Level (n50)
PlanetLab (n50)
BRITE (n50)
0 2 3 5 11 22
0 2 3 5 11 22
0 2 3 5 11 22
k
k
k

• Common pattern is not good

If your neighbors have the same wiring pattern,
it pays to be selfish!
20
Connecting on a Best Response graph
AS-Level (n50)
PlanetLab (n50)
BRITE (n50)
0 2 3 5 11 22
0 2 3 5 11 22
0 2 3 5 11 22
k
k
k

• The BR graph is highly optimized!

If your neighbors are selfish, it is OK to be
naïve!
21
SNS vs. Heuristics Social Cost

• Macroscopic view
• Focusing on the social welfare

(k2) k-Random/BR k-Closest/BR k-Regular/BR
BRITE 1.44 1.53 3.61
PlanetLab 2.23 1.48 3.84
AS 2.04 1.90 4.78
The network is better off with selfish nodes!
22
Real-Time Applications

• Min-Max Best Response
• Worst delay
• in the overlay

0 2 3 5 11 22
k
23
SNS with Variable Degree

• Real-time applications
• Variable degree
• through LS
• Swap 1 link
• Add 1 link
• Drop 1 link

100 links
120 links
Application requirement
(Performance when k5, n50 i.e. 250 links)
24
Implications to Overlay Routing
Selfish Neighbor Selection
Implications to File Sharing
Implications to Service Provisioning
25
Basic design of EGOIST Link state
protocol Measurements of distance to candidate
neighbors Wirings according to chosen strategy
Re-wirings every T second A newcomer bootstraps
by connecting to arbitrary neighbors
26
EGOIST Performance
Best Response
27
EGOIST Passive Measurements

• Passive measurements based on virtual
• coordinates (pyxida system) with minimal
• cost

28
EGOIST Other Metrics

• End-to-end available bandwidth (pathchirp) with
  minimal measurement overhead
• CPU load (loadavg)

29
EGOIST Marginal Utility of Rewiring
Lazy BR (threshold 10)
BR

• There exists a performance knee (k3 or 4)
• Re-wirings could be reduced with lazy BR

30
EGOIST Effect of Churn
Efficiency Index
Connectivity quality

• Connectivity is guaranteed (in T/n time)
• HybridBR (a connected ring is maintained)
• delivers much of the efficiency of BR

31
EGOIST Effect of Churn
Efficiency Index
Connectivity quality

• BR and Hybrid BR dominate all the other
• heuristics
• HybridBR pays off at high churn

32
EGOIST Other Work

• CPU and memory load is very low
• Robust to cheating
• Scalability
• via topological sampling
• via layered architecture
• Applications including multi-player P2P games,
  real-time traffic over IP etc.

33
Implications to Overlay Routing
Selfish Neighbor Selection
Implications to File Sharing
Implications to Service Provisioning
34
Modern File Sharing Systems

• Parallel upload/ download
• - Swarming
• Local scheduling
• - Local Rarest First
• Flat connectivity
• - Choke/unchoke

Internet
Seeder
Transit ISP
Transit ISP
Access ISP
Access ISP
Access ISP
Leecher
Overlay node
35
n-way Broadcast

• Synchronization
• - Distributed databases
• - Backups
• Batch parallel processing
• - The files have to be received by all nodes
  before the next step
• of processing begins

Internet
36
Preliminary Solutions

• n co-existing swarms
• (-) Stress of physical links
• (-) Exchange of multiple chunks in parallel
  overpartitions
• the uplink capacity Tian et al.,
  ICPP06
• End-system multicast (mesh) SplitStream, Bullet
  
• (-) Creates an overlay for each swarm
• (-) No coordination among swarms
• (-) Monitor overhead

37
Design Strategies for n-way Broadcast

• Joint optimization of upload/download
• while participating in many swarms
• Data Agnostic
• - Keeps swarming and local scheduling
• Bandwidth-Centric
• - Max-flow to approximate swarming behavior
• Massoulie et al., Infocom07
• Bounded Degree

38
Reducing the Average Download Time

• Objective Minimize the average download time
• Max-Sum
• Neighbor selection strategy of node vi
• max (sum (MaxFlow(vi, vj)), for all vj

39
Reducing the Download Time

• Objective Minimize the total download time
• Max-Min
• Neighbor selection strategy of node vi
• max (min (MaxFlow(vi, vj)), for all vj

40
Optimized Graphs and Swarming

• Formation of stable graphs
• Each node strives to improve both the
• upload and download flow
• Performance of swarming on optimized graphs
• - Max flow might not be realizable

41
Performance Evaluation
Naive
Max-Sum
Max-Min
Node ID
Delivery Time
Selfish Upload Protects the uplink capacity of
the slow node ? Improves the download time in the
system
File ID
File ID
File ID

• Flattens distribution time!
• Guarantees synchronization!
• Comparable average download time

42
Other Work File Searching

• Best response max nodes reached

4
Bootstrap Server
1
6
3
5
2
selfishly
TTL of scoped flooding is 2
? Maximum Coverage Problem
43
Implications to Overlay Routing
Selfish Neighbor Selection
Implications to File Sharing
Implications to Service Provisioning
44
Server Selection
Hardware server
45
Centralized Deployment
Generic Service Host
Software server
Demand change e.g. Flash crowd,
time-of-day effect
46
Dynamic Service Deployment
Generic Service Host
Software server
Demand change e.g. Flash crowd,
time-of-day effect
47
Distributed Service Migration (DSM)

• Solve k-median or UFL
• in an r-ball
• ..BUT nodes outside the r-ball are totally
  neglected

• Iterate until
• convergence

48
DSM Properties

• Convergence
• Migration only if the cost of facilitating
  the demand decreases at least be a, converges in
  O(log1a n) steps
• We can control the speed of convergence by tuning
  a
• Limited horizon view requirement
• r regulates the trade-off between scalability and
  performance

49
DSM Evaluation

• Similar results for UFL under different
• cost functions to open and maintain the server

50
Dynamic vs. Static Deployment

• Static
• deployment

DSM
DSM

• Dynamic deployment

51
Conclusions

• What is the performance gain that can be achieved
  by a selfish node?
• ? Selfish nodes can reap substantial performance
  gain.

• What is the impact of selfish neighbor
• selection to overlay network performance?
• ? Surprisingly, the evolving graphs have also
  good performance!

52
Conclusions

• What are the implications of selfish neighbor
  selection to system design?
• Selfish wiring strategies are easily realizable
• Selfish wiring behavior can be used towards
  distributed overlay network creation and
  maintenance
• Selfish wiring must be a component of any system
  to protect it from abuse
• Selfish wiring behavior can be used for efficient
  dynamic service provisioning

53
Thank You! http//csr.bu.edu/snshttp//csr.bu
.edu/dfl