An Efficient and Fault-Tolerant Update Commitment Protocol for Weakly Connected Replicas

Abstract

Abstract. Mobile and other loosely-coupled environments call for decentralized optimistic replication protocols that provide highly available access to shared objects, while ensuring eventual convergence towards a strongly consistent state. In this paper we propose a novel epidemic weighted voting protocol for achieving such goal. Epidemic weighted voting approaches eliminate the single point of failure limitation of primary commit approaches. Our protocol introduces a significant improvement over other epidemic weighted voting solutions by allowing multiple, happened-before related updates to be committed at a single distributed election round. We demonstrate that our protocol is especially advantageous with the weak connectivity levels that characterize mobile and other loosely-coupled networks. We support such claims by presenting comparison results obtained from side-by-side execution of reference protocols in a simulated environment.

Exploiting Our Computational Surroundings for Better Mobile Collaboration

“…We believe these requirements are desirable for most collaborative applications, as they allow applications to access a strongly consistent view, in complement to the tentative view [16]. They are verified in a number of representative systems such as Bayou [16], IceCube [10], Coda [12], CVS [3], Deno [9] or VVWV [1].…”

“…The evaluation used a C# implementation of the VVWV protocol [1], extended as proposed in Section 2, in a simulator. 4 The simulator includes a collection of mobile nodes, randomly distributed by a set of network partitions.…”

“…However, if we decouple the agreement phase from the remaining phases, we observe that proposed agreement algorithms (e.g. Bayou's primary-commit [16] or VVWV's weighted voting [1] agreement algorithms) do not actually require access to all the information an update holds; namely, they are not concerned with the op u and precond u components of updates in a schedule. Instead, they may simply deal with some meta-data representation that identifies the (i) updates, (ii) their order, and (iii) their state (i.e.…”

“…Second, propagation of updates ensures the prefix property [16]: if a replica r holds an update u x that was initially generated at another replica x, then r will also have received all other updates held by x (in sch x ) prior to u x . For example, Bayou [16], Coda [12], and VVWV [1] satisfy the above properties.…”

A mechanism for replicated data consistency in mobile computing environments

“…In order to ensure replicated data consistency in mobile computing environments, several approaches relax one-copy serializability. However, most of these approaches require intensive message exchange [2,5,9], or, that transaction operations (stored in log), be re-executed several times. For example, in [11] the servers may need to undo previously executed "tentative write" effects and re-execute them in another order.…”