Managing Dynamic Shared State

1
Managing Dynamic Shared State

• Jiyuan Wang

2
Dynamic Shared State

• What is it?
• The changing information that multiple machines
  must maintain about the N-VE.
• Information included participants, virtual
  world, relations between the participants,
  relations between participants and virtual world.
• Why need to do dynamic shared state management?
• It provides the common context that makes the
  virtual environment into a collaborative or
  competitive experience.

3
Consistency-Throughput Tradeoff
Scenario I
Joe
Mary
Im at(10,20)
Currently
Im at(15,25)
After 100 ms
100 ms is the network latency
4
Consistency-Throughput Tradeoff
Scenario II
Joe
Mary
Im at(10,20)
Currently
Transmit
Joe is at(10,20)
Acknowledge
After 100 ms
Im at(15,25)
After 200 ms
5
Consistency-Throughput Tradeoff

• It is impossible to allow dynamic shared state to
  change frequently and guarantee that all hosts
  simultaneously access identical versions of that
  state.
• Available network bandwidth must be allocated
  between messages for updating the dynamic state
  and messages for maintaining a consistent view of
  that dynamic shared state among participants in
  the N-VE.

6
Spectrum of Dynamic Shared State Management
Determined by the Consistency-Throughput Tradeoff
7
Centralized Information Repositories

• File Repository
• Repository in Server memory
• --pull
• --push
• Virtual Repositories
• --virtual central repository
• --client consistency selection

8
File Repository
User
User
Synchronization Locks by NFS/AFS
User
User
User
User
Centralized Data Store
Update
Read
State
State
Read
State
State
Update
9
Problem with File Repository

• Performance is slow.
• Opening remote file is a slow procedure
• Not easy to support many users.
• More users?More files?Longer to open each file

10
Repository in Server Memory
User
Synchronization Locks by
User
User
User
User
User
Centralized Data Store
Per-client FIFO Event Queues
Per-client FIFO Event Queues
State
State
Update
Read
State
State
Read
Update
11
Advantages and Problems

• Advantages
• Keep shared states in server memory
• No need to perform open and close operation on
  files
• Server process arbitrate among incoming requests
• Server support batched operations
• Problems
• Unreliable
• Operating system resource consuming

12
Systems Using Server Memory Repository

• NEC Network VR Project
• pull information
• Server maintains a FIFO queue for each client
• Clients request can cause delay
• Shastra SystemDeveloped at Purdue University
• push information
• Server uses round-robin token-passing scheme

13
Virtual Repositories
A consistency protocol can establish a virtual
central state repository by ensuring reliable
message delivery, effective message
dissemination, and message ordering.
14
Virtual Repositories

• Advantage
• Eliminates the performance bottleneck
• Eliminate the bandwidth bottleneck
• Better fault tolerance
• Systems using virtual repositories
• DIVE
• BrickNet
• Client Consistency Selection

15
Characteristics of Central Repository

• Easy programming model
• Guarantee the information consistency
• Flexibility on shared state update
• Data access and update operations require an
  unpredictable amount of time to complete
• Require considerable communications overhead
• Limited number of simultaneous users, 50-100

16
Frequent State Regeneration

• For high shared state update rate
• Blind broadcasting
• Explicit Entity ownership
• Prevent update conflict

17
Scenario of Ownership
Update ball position per Joe
Proxy update
Request update ball position
Update ball position per Mary
Notify lock transfer
Ownership transfer
Acknowledge lock transfer
Update ball position per Joe
Request ball ownership
Grant ball ownership
Update ball position per Mary
18
Systems Using Frequent State Regeneration

• Dogfight
• Doom, Diablo and Interstate 76
• Telesurgery
• Video Conference

19
Advantages and Drawbacks

• Advantages
• Easy to add multi-user capabilities
• .need no server,consistency and even lock manager
• Support large number of users
• Drawbacks
• High network bandwidth utilization
• Network latency and network jitter
• Differences between local and remote entities