This paper presents the design, implementation, and evaluation of the replication framework of Deno, a decentralized, peer-to-peer object-replication system targeted for weakly connected environments. Deno uses weighted voting for availability and pair-wise, epidemic information flow for flexibility. This combination allows the protocols to operate with less than full connectivity, to easily adapt to changes in group membership, and to make few assumptions about the underlying network topology. We present two versions of Deno's protocol that differ in the consistency levels they support. We also propose security extensions to handle a class of malicious actions that involve misrepresentation of protocol information. Deno has been implemented and runs on top of Linux and Win32 platforms. We use the Deno prototype to characterize the performance of the Deno protocols and extensions. Our study reveals several interesting results that provide fundamental insight into the benefits of decentralization and the mechanics of epidemic protocols.

Managing failures and configuring systems properly are of critical importance for robust distributed services. Unfortunately, protocols offering strong fault-tolerance guarantees are generally too costly and insensitive to performance criteria. Yet, system management in practice is often ad-hoc and ill-defined, leading to under-utilized capacity or adverse effects from poorly-behaving machines. This paper proposes a new abstraction called linkattestation groups (LA-Groups) for building robust distributed systems. Developers specify application-level correctness conditions or performance requirements for nodes. Nodes vouch for each other's acceptability within small groups of nodes through digitally-signed link attestations, and then apply a link-state protocol to determine these group relationships.

This paper introduces and analyzes a variant of distributed gossip which is motivated by the sharing of recommendations in a social network. The social settings bear two implications on gossip. First, rumors fade after a few hops, and so does our gossip mechanism. Second, users require a rumor to be substantiated by multiple, independent sources in order to adopt it. Consequently, in our social gossip a message is adopted only when it is received over a threshold of independent paths. Social gossip is a new, highly relevant and practically motivated variant of distributed gossip, whose analysis contributes to the fundamental theory of distributed algorithms.

We consider the problem of disseminating an update known to a set of servers to other servers in the system via a gossip protocol. Some of the servers can exhibit malicious behavior.

Abstract. Network overlays have been the subject of intensive research in recent years. The paper presents an overlay structure, S-Fireflies, that is self-stabilizing and is robust against permanent Byzantine faults. The overlay structure has a logarithmic diameter with high probability, which matches the diameter of less robust overlays. The overlay can withstand high churn without affecting the ability of active and correct members to disseminate their messages. The construction uses a randomized technique to choose the neighbors of each member, while limiting the ability of Byzantine members to affect the randomization or to disturb the construction. The basic ideas generalize the original Fireflies construction that withstands Byzantine failures but was not self-stabilizing. 1

Abstract. Data redundancy in DHTs is commonly accomplished through automatic replication of values to a set of close participating nodes. Such copies, once written, should not be modified, as there is no inherent DHT function that can operate on a dynamic set of mutable replicas. In this paper, we present XOROS — a system based on the Kademlia routing scheme, that addresses the problem by implementing a Byzantinetolerant protocol for serializable data updates directly at the peer-to-peer level. Based upon related works that study distributed replica synchronization, mutual exclusion and communication in the presence of Byzantine behavior, we propose a unified DHT-based algorithm, that ties corresponding practices together in order to consistently propagate changes to all primary replicas of any key-value pair stored in the network. A multitude of applications may benefit from the resulting distributed read/write storage substrate as it retains all the advanced features of DHTs and is backwards compatible with existing put/get semantics. 1

Rumor spreading algorithms are a useful way to disseminate information to a group of players in the presence of faults. Rumors are either spread by pushing, in which the players knowing the rumor call other players at random and spread the rumor, or by pulling, where players who do not know the rumor call other players and ask for any new rumors.

In this paper, we present an infrastructure for providing secure transactional replication support for peer-to-peer, decentralized databases. We first describe how to effectively provide protection against external threats, malicious actions by servers not authorized to access data, using conventional cryptography -based mechanisms. We then classify and present algorithms that provide protection against internal threats, malicious actions by authenticated servers that misrepresent protocol-specific information. Our approach to handling internal threats uses both cryptographic techniques and modifications to the update commit criteria. The techniques we propose are unique in that they not only enable a tradeoff between performance and the degree of tolerance to malicious servers, but also allow for individual servers to support non-uniform degrees of tolerance without adversely affecting the performance of the rest of the system. We investigate the cost of our securi

Overlay networks provide important routing functionality not easily supported directly by the Internet. Distributed Hash Tables (DHTs) have been proposed to support such overlay networks. While it is often straightforward to support overlay networks on DHTs, this choice can be questioned. DHTs dictate routes that are not optimal, and DHTs are hard to secure. As overlay networks are beginning to be deployed for critical applications, efficiency and security are becoming important attributes. We present an alternative support structure called Fireflies. Fireflies provides each of its members with a complete view of its live peers. A small subset of these peers are marked as neighbors. With high probability, the mesh formed by the members and their neighbor links has a diameter logarithmic in the number of live members, and connects all the reachable members that are not Byzantine. Fireflies uses several randomized algorithms, which are briefly described. We also discuss how Fireflies may be used to build intrusion-tolerant overlay networks. 0 We are supported by DARPA’s SRS program, a MURI

Abstract not found