Managing Dynamic Shared State

Introduction

NVE

Shared State

Problem : Latency

Consistency-Throughput Tradeoff

High Consistency versus Highly Dynamic

Tradeoff Spectrum

3 Technics

Centralized Shared Repositories

Shared File Directory

Repository in Server Memory

Central server : Pull

Central server : Server Push the information

Virtual Repositories

Advantage and Drawback of Shared Repositories

Conclusion for Shared Repositories

Frequent State Regeneration / Blind Broadcasts

Explicit Object Ownership

Lock Manager

Update by Non-owner: Proxy Update

Update by Non-owner : Ownership Transfer

Reducing Broadcast Scope

Advantages of Blind Broadcasts

Disadvantages of Blind Broadcasts

Dead Reckoning

Dead Reckoning Issues

Prediction Techniques

Hybrid polynomial prediction

PHBDR (Position History-Based Dead Reckoning)

Limitations of Derivative Polynomials

Object-Specialized Prediction

The Phugoid model

Convergence Algorithms

Linear convergence

Higher Level Convergence

Non-regular update packet transmission

Non-regular update packet transmission (2)

Dead Reckoning Conclusion

Conclusion

eScience