Architectural support for designing fault-tolerant open distributed systems

Abstract

Permanence: The results of committed transactions will not be lost.Serializability: The results of executing transactions concurrently are the same as if they were executed serially.Use of the transaction concept to model distributed computations provides a convenient means to solve the concurrency control and redundancy management problems.' The concurrency control problem consists of three tasks: assigning an order to all transactions, identifying conflicting transactions, and synchronizing transactions to resolve the identified conflicts. Basically, there are three approaches to concurrency control: time-stamp-based schemes, locking protocols, and optimistic techniques.2

Take Intelligent Risk and Optimize Decision Based on Time, Available Resources and Risk Tolerance Limits

“…Therefore, voting is widely used in consistency and agreement algorithms in distributed systems [23]. Many voting protocols have been studied elsewhere by the research community under various application settings and environments [6,10,25]. [20] gives a very good summary of various issues in voting.…”

Expected Time for Obtaining Dependable Data in Real-Time Environment

“…We further assume that C(t) is uniformly distributed over the interval [0, T 2 ] and V(t) = 1 3 . From (5), we can derive the probability of getting a valid data before time t:…”

“…Therefore, voting is widely used in consistency and agreement algorithms in distributed systems [18] Many voting protocols have been studied elsewhere by the research community under various application settings and environments [3,5,19]. [14] gives a profound summary of various issues in voting.…”

Group communication in the Amoeba distributed operating system

“…Because NVRAM is a reliable medium, this implementation provides the same degree of fault tolerance as the other implementations, while the performance is much better. A similar optimization has been used in [10,24,14].…”