Scalable Peer-to-Peer Networked Virtual Environment - Multimedia ...

yardbellAI and Robotics

Nov 14, 2013 (3 years and 6 months ago)

55 views

2004/07/01

1

Scalable Peer
-
to
-
Peer

Networked Virtual Environment


Shun
-
Yun Hu* & Guan
-
Ming Liao
+


* Dept. of CSIE, Tamkang Univ.

+
Institute of Physics, Academia Sinica


2004/07/01

2004/07/01

2

What is Networked Virtual
Environment (NVE)?


Virtual Reality + Internet



People (avatar), objects, terrain, agents



Military simulations (’80)


Online Games (mid
-
‘90)



Eventual goal: World Wide Web
-
like


(global interactive space)

2004/07/01

3


2004/07/01

4

The NVE Problem


Many real people (100 ~ 10,000)


Want to interact in a VE (
via PC & Internet
)


How do they see each other properly?


1.
Consistency

2.
Performance / responsiveness

3.
Security / fairness

4.
Scalability

5.
Persistency

6.
Reliability / fault
-
tolerance

2004/07/01

5

A simple solution

N * (N
-
1) connections
≈ O(N
2
)


Not scalable!

Source: [Funkhouser95]

2004/07/01

6

A better solution (client
-
server)

N connections = O(N)


server is bottleneck

Source: [Funkhouser95]

2004/07/01

7

Current solution (server
-
cluster)

Still limited by servers. Expansive to deploy & maintain
.

Source: [Funkhouser95]

2004/07/01

8

The Scalability Problem


Observed in many natural & artificial systems



Two characteristics (resource
-
related)


Joinability


Maintainability



Strategies to achieve scalability


Resource
-
growing


Decentralized resource consumption at end
-
point

2004/07/01

9

Existing Solutions


Increase Resource


More servers!


Complexity, cost, maintenance,
centralization



Decrease Consumption


Packet compression & aggregation


Interest Management


(Area of Interest, AOI)

2004/07/01

10

Promise of Peer
-
to
-
Peer (P2P)

[Stoica et al. 2003]


Distributed systems without any centralized control
or hierarchical organization


Runs software with equivalent functionality




Famous examples


File
-
sharing:


Napster, Gnutella, eDonkey


Parallel Computing:

SETI@Home (UC Berkeley)


2004/07/01

11

A P2P overlay network





[Keller & Simon 2003]

2004/07/01

12

Challenges of P2P NVE


Correct topology maintenance


Fully
-
connected


Consistent view



Efficient content retrieval


Messaging with only relevant (AOI) neighbors



Neighbor Discovery Problem

2004/07/01

13

Voronoi
-
based P2P NVE
Overview


Locality of interest



Connections to neighbors within AOI


Mutual cooperation in neighbor discovery


Minimal set of neighbors maintained

2004/07/01

14

Voronoi Diagram


Plane partitioned into
regions

by
sites
, such that the
region contains all points closest to its site

2004/07/01

15

Voronoi Diagram


Properties of Voronoi Diagram


Can be used to find
k
-
nearest neighbor

easily


Can be constructed in O(n log n)

2004/07/01

16

Design Concepts


Use Voronoi to solve
Neighbor Discovery Problem


Identify
enclosing

and
boundary

neighbors


Each node constructs a Voronoi of its neighbors


Mutual collaboration in neighbor discovery

Circle
: Area of Interest (AOI)

White: self

Yellow: enclosing neighbor

Pink: enclosing & boundary

Green: other neighbor

L. blue: boundary neighbor

D. blue: unknown neighbor

2004/07/01

17

Procedure


Join


Send join request to any existing node


Join request forwarded to
acceptor region


Acceptor region
’s host sends back its neighbor list


Joining host

contacts those neighbors individually

2004/07/01

18

Procedure


Movement


Send movement to all neighbors, mark
boundary neighbors


Neighbors check for AOI & enclosing neighbors overlap


Connect to new neighbor upon notification (update Voronoi)


Disconnect any non
-
overlapped neighbor

2004/07/01

19

Procedure


Other actions (chat, jump, trade)


Send message to individual or to all neighbors


Leave


Simply disconnect


Others update Voronoi (query new boundary)

2004/07/01

20

Analysis of Design

Consistency


Enclosing neighbors


fully connected & consistent

Performance / Responsiveness


Direct connection (no server relay)

Security

Scalability


Resource
-
growing &
decentralized resource consumption

Persistency

Reliability


Self
-
repair for small number of node failures

2004/07/01

21

Implementation (in
-
progress)


Fortune’s sweepline algorithm for Voronoi


High Level Architecture (HLA) Run
-
time
infrastructure (RTI) interface



Open source NVE development platform



2004/07/01

22

Preliminary Results


Average message size received per node
0
500
1000
1500
2000
2500
3000
10
30
50
70
90
110
Number of Nodes
Message Size (Bytes)
2004/07/01

23

Summary


Analyze scalability problem & potential solution



A promising solution: Voronoi
-
based P2P NVE


Leverage knowledge of each peer to maintain topology



Properties


Simple: Voronoi maintenance, no centralized processing


Efficient: direct transmission, minimal topology overhead


Ideal interest management: mostly AOI messages


2004/07/01

24

Future Works


Variable size AOI


Persistent state management in P2P


Security


Combination with 3D streaming



Use in education & small community


2004/07/01

25

Acknowledgements


WISE Lab, Tamkang Univ.


Joaquin Keller (Solipsis)


LSCP, Institute of Physics, Academia Sinica



Prof. Wen
-
Bing Horng


Dr. Chin
-
Kun Hu



Prof. Jiung
-
yao Huang

2004/07/01

26

Q & A



Thank You!

2004/07/01

27

Related Works (1)

[Knutsson et al. 2004] (Univ. of Penn.)



Pastry + Scribe


Coordinators





Overlay overhead

2004/07/01

28

Related Works (2)

[Kawahara et al. 2004] (Univ. of Tokyo)



Fully
-
distributed


Nearest
-
neighbor


List exchange




High transmission


Overlay partition

2004/07/01

29

Related Works (3): Solipsis

[Keller & Simon 2003] (France Telecomm R&D)



Fully
-
distributed


Links with AOI neighbor


Mutual cooperation


Inside convex hull




Inconsistent topology