Abstract
Communication of large data volumes is a core functionality of distributed systems middleware, namely, for interconnecting components, for distributed computation and for fault tolerance. This common functionality is however achieved in different middleware platforms with various combinations of operating system and application level protocols, both standardized and ad hoc, and including implementations on managed runtime environments such as Java. In this paper, in contrast with most previous work that focus on performance, we point out that architectural and implementation decisions have an impact in throughput stability when the system is heavily loaded, precisely when such stability is most important. In detail, we present an experimental evaluation of several communication protocol components under stress conditions and conclude on the relative merits of several architectural options.

Abstract
Most proposed gossip-based systems use an ad-hoc design. We observe a low degree of reutilization among this proposals. We present how this limits both the systematic development of gossip-based applications and the number of applications that can benefit from gossip-based construction. We posit that these reinvent-the-wheel approaches poses a significant barrier to the spread and usability of gossip protocols. This paper advocates a conceptual design framework based upon aggregating basic and predefined building blocks (B 2). We show how to compose building blocks within our framework to construct more complex blocks to be used in gossip-based applications. The concept is further depicted with two gossip-based applications described using our building blocks.

Abstract

Abstract

Abstract
Publish/subscribe mechanisms for scalable event dissemination are a core component of many distributed systems ranging from Enterprise Application Integration middleware to news dissemination in the Internet. Hence, a lot of research has been done on overlay networks for efficient decentralized topic-based routing. Specifically, in gossip-based dissemination, bringing nodes with shared interests closer in the overlay makes dissemination more efficient. Unfortunately, this usually requires fully disclosing interests to nearby nodes and impacts reliability due to clustering. In this paper we address this by starting with multiple overlays, one for each topic subscribed, that then separately self-organize to maximize the number of shared physical links, thereby leading to reduced message traffic and maintenance overhead. This is achieved without disclosing a node's topic subscription to any node that isn't subscribed to the same topic and without impacting the robustness of the overlay. Besides presenting the overlay management protocol, we evaluate it using simulation in order to validate our results. © IPTPS 2010.All right reserved.

Abstract
Multi-master update everywhere database replication, as achieved by protocols based on group communication such as DBSM and Postgres-R, addresses both performance and availability. By scaling it to wide area networks, one could save costly bandwidth and avoid large round-trips to a distant master server Also, by ensuring that updates are safely stored at a remote site within transaction boundaries, disaster recovery is guaranteed. Unfortunately, scaling existing cluster based replication protocols is troublesome. In this paper we present a database replication protocol based on group communication that targets interconnected clusters. In contrast with previous proposals, it uses a separate multicast group for each cluster and thus does not impose any additional requirements on group communication, easing implementation and deployment in a rea setting. Nonetheless, the protocol ensures one-copy equivalence while allowing all sites to execute update transactions. Experimental evaluation using the workload of the industry standard TPC-C benchmark confirms the advantages of the approach.