Efficient Triangulation for P2P Networked Virtual Environments - - PowerPoint PPT Presentation

▶
efficient triangulation for p2p networked virtual
SMART_READER_LITE
LIVE PREVIEW

Efficient Triangulation for P2P Networked Virtual Environments - - PowerPoint PPT Presentation

Jan 13, 2023 168 likes •407 views

Efficient Triangulation for P2P Networked Virtual Environments Eliya Buyukkaya, Maha Abdallah CS Department (LIP6) University of Paris 6 Networked Virtual Environment Features A way of communication & interaction Immersion in

slide-1
SLIDE 1

Eliya Buyukkaya, Maha Abdallah CS Department (LIP6) University of Paris 6

Efficient Triangulation for P2P Networked Virtual Environments

slide-2
SLIDE 2

NetGames 2008 Maha Abdallah 2

Networked Virtual Environment

Features

A way of communication & interaction Immersion in a synthetic world through avatars Perceived shared space & time

Applications

Military training (since 80’s) Remote teaching & education, online meetings, … Massively Multiplayer Online Games (MMOGs)

slide-3
SLIDE 3

NetGames 2008 Maha Abdallah 3

NVEs : Current state

Multi-billion dollar industry Growing number of subscribers

10 million subscribers for World of Warcraft (WoW) 800,000 concurrent players for WoW Burning Crusade

However…

≤ 5000 concurrent players per world Why & How to scale ?

slide-4
SLIDE 4

NetGames 2008 Maha Abdallah 4

Towards a 3D WEB ?

Ultimately… Can we have all users concurrently share

  • ne huge, realistic, and immersive

virtual world ?

slide-5
SLIDE 5

NetGames 2008 Maha Abdallah 5

The scalability issue

slide-6
SLIDE 6

NetGames 2008 Maha Abdallah 6

System Models for NVEs (1)

Early stage : Client/server

Server represents…

  • a bottleneck
  • a single point of failure

Does NOT scale

slide-7
SLIDE 7

NetGames 2008 Maha Abdallah 7

System Models for NVEs (2)

Current deployments: server cluster

Expensive Limited scalability Crowding problems

slide-8
SLIDE 8

NetGames 2008 Maha Abdallah 8

System Models for NVEs (3)

P2P as a promising technology

Highly scalable Cost effective

slide-9
SLIDE 9

NetGames 2008 Maha Abdallah 9

P2P challenges and Issues

neighbor discovery

&

self reorganization neighborhood relationships

Overlay topology structuring

Arbitrary Application semantics-oriented

Dynamic topology maintenance

Users’ join/leave Users’ continuous movement in NVEs

Distributed content management

slide-10
SLIDE 10

NetGames 2008 Maha Abdallah 10

Related Works (1)

  • SimMUD

SimMUD [Knutsson et al. 2004]

  • Region-based
  • Pastry/Scribe-based
  • Multicast groups
  • Fixed-size regions
  • Coordinator-based
  • Message relay

[Iimura et al. 2004]

  • Similar w/ no message relay

Source: [ Knutsson et al. 2004]

Performance & scalability problems

slide-11
SLIDE 11

NetGames 2008 Maha Abdallah 11

Related Works (2)

[Kawahara et al. 2004]

  • Connection to closest neighbors
  • Regular neighbor list exchange
  • High communication cost
  • MOPAR

MOPAR [Yu et al. 2005]

  • Hexagonal cells
  • 1 master/cell
  • Regular master neighbor list exchange
  • High communication cost

Source: [ Yu et al. 2005]

slide-12
SLIDE 12

NetGames 2008 Maha Abdallah 12

Related Works (3)

  • Solipsis

Solipsis [Keller et al. 2003]

  • Connections to all AOI neighbors
  • Mutual cooperation for neighbor discovery
  • Incomplete neighbor discovery
  • Possible inconsistent topology
  • Crowding issues => high communication cost
  • VON

VON [Hu et al. 2004]

  • Voronoi-based
  • Connection to all AOI neighbors
  • Mutual cooperation for neighbor discovery
  • Crowding issues => high communication cost
  • High overlay maintenance cost

Source: [ Keller et al. 2003] Source: [ Hu et al. 2004]

slide-13
SLIDE 13

NetGames 2008 Maha Abdallah 13

Related Works (4)

  • VSM

VSM [Hu et al. 2008], [Buyukkaya et al. 2008]

  • VON-based
  • Distributed state management
  • Crowding issues => high communication cost
  • High overlay maintenance cost
  • VoroCAST

VoroCAST [Hu et al. 2008]

  • VON-based
  • Multicast within AOI
  • Reduced bandwidth consumption per node
  • Higher latency
  • High overlay maintenance cost

Source : [Buyukkaya et al. 2008] Source : [Hu et al. 2008]

slide-14
SLIDE 14

NetGames 2008 Maha Abdallah 14

[Varvello et al. 2007]

  • Delaunay-based
  • Cluster creation when crowding
  • Stretching distances within cluster
  • All Delaunay flip operations are still performed

Related Works (5)

Source : [Varvello et al. 2007]

High overlay maintenance cost

slide-15
SLIDE 15

NetGames 2008 Maha Abdallah 15

Voronoi/Delaunay Voronoi/Delaunay : the right choice ? : the right choice ?

The BIG Question The BIG Question…

slide-16
SLIDE 16

NetGames 2008 Maha Abdallah 16

A Delaunay triangulation for a set of vertices A in a 2-D plane is a triangulation DT(A) such that no vertex of A is inside the circumcircle

  • f any triangle in DT(A)

Features

Balanced triangulation Proximity-based => Locality Low number of neighbors => scalability Dual graph of Voronoi diagram => state management

Delaunay Triangulation for NVEs

slide-17
SLIDE 17

NetGames 2008 Maha Abdallah 17

Issue

  • High connection change rate (edge-flipping)
  • High maintenance cost in NVEs & MMOGs

Perfect Delaunay Triangulation in NVEs necessary

  • r sufficient

?

Delaunay Triangulation for NVEs

Edge-flipping 02 replaced by 14

Before 4’s movement After 4’s movement

slide-18
SLIDE 18

NetGames 2008 Maha Abdallah 18

Delaunay Revisited…

Flip Flip-

  • free

free area area

Area where the node can move without triggering edge-flipping

Standard Delaunay

Basic Idea Basic Idea

Reduce communication cost by maximizing a node’s flip-free area while maintaining a valid triangulation

slide-19
SLIDE 19

NetGames 2008 Maha Abdallah 19

Delaunay Revisited…

Standard Delaunay Our variation Maximizing flip-free region

Stretched flip Stretched flip-

  • free

free area area the region composed of all the triangles for which a node is a vertex.

slide-20
SLIDE 20

NetGames 2008 Maha Abdallah 20

Delaunay Revisited: node movement

Is flip-free area convex ? YES

Flip only when crossing the base of one of its triangles

0 crosses base 23 Edge-flipping 23 replaced by 05

Increased degree ?

Flip w/ triangles having longest common edge

slide-21
SLIDE 21

NetGames 2008 Maha Abdallah 21

Delaunay Revisited: flip operations

Is flip-free area convex ? NO

Flip only when => crossing the base of a triangle => Triangle anomalies

  • ccur

Edge-flipping 04 replaced by 13 0 moves towards 2

slide-22
SLIDE 22

NetGames 2008 Maha Abdallah 22

Delaunay Revisited: node insertion

n entering triangle T

1.

check neighbor triangles w/ common edge with T

2.

If n inside one of the circumcercles

  • Flip with the corresponding triangle(s)

Entry point

8 enters through node 0

slide-23
SLIDE 23

NetGames 2008 Maha Abdallah 23

Conclusions

Is Delaunay/Voronoi the best choice for NVEs ? Consider weakening conditions/alternatives

Imperfect Delaunay is enough What impact on corresponding Voronoi ?

  • Region management, load balancing, etc.

Voronoi dual graph not ideal for dynamic state management

Current work

Formal analysis Extensive evaluations w/ real traces (MMOGs and social NVEs)

  • Latency vs message cost
  • Packets/sec

Decorrelate state/object management from virtual position Dual overlay !