John Gilmore

A State Management and Persistency Architecture for Peer-to-Peer Massively Multi-user Virtual Environments

Recently, there has been significant research focus on Peer-to-Peer (P2P) Massively Multi-user Virtual Environments (MMVEs). A number of architectures have been presented in the literature to implement the P2P approach. One aspect that has not received sufficient attention in these architectures is state management and state persistency in P2P MMVEs. This work presents and simulates a novel state management and persistency architecture, called Pithos.
In order to design the architecture, an investigation is performed into state consistency architectures, into which the state management and persistency architecture should fit. A novel generic state consistency model is proposed that encapsulated all state consistency models reviewed. The requirements for state management and persistency architectures, identified during the review of state consistency models, are used to review state management and persistency architectures currently receiving research attention.

Identifying some deficiencies present in current designs, such as lack of fairness, responsiveness and scalability, a novel state management and persistency architecture, called Pithos, is designed. Pithos is a reliable, responsive, secure, fair and scalable distributed storage system, ideally suited to P2P MMVEs. Pithos is implemented in Oversim, which runs on the Omnet++ network simulator. An evaluation of Pithos is performed to verify that it satisfies the identified requirements.

It is found that the reliability of Pithos depends heavily on object lifetimes. If an object lives longer on average, retrieval requests are more reliable. An investigation is performed into the factors influencing object lifetime. A novel Markov chain model is proposed which allows for the prediction of objects lifetimes in any finite sized network, for a given amount of redundancy, node lifetime characteristics and object repair rate.