Many distributed systems work on a common shared state; in such systems, distributed agreement is necessary for consistency. With an increasing number of servers, these systems become more susceptible to single-server failures, increasing the relevance of fault-tolerance. Atomic broadcast enables fault-tolerant distributed agreement, yet it is costly to solve. Most practical algorithms entail linear work per broadcast message. AllConcur - a leaderless approach - reduces the work, by connecting the servers via a sparse resilient overlay network; yet, this resiliency entails redundancy, limiting the reduction of work. In this paper, we propose AllConcur+, an atomic broadcast algorithm that lifts this limitation: During intervals with no failures, it achieves minimal work by using a redundancy-free overlay network. When failures do occur, it automatically recovers by switching to a resilient overlay network. In our performance evaluation of non-failure scenarios, AllConcur+ achieves comparable throughput to AllGather - a non-fault-tolerant distributed agreement algorithm - and outperforms AllConcur, LCR and Libpaxos both in terms of throughput and latency. Furthermore, our evaluation of failure scenarios shows that AllConcur+'s expected performance is robust with regard to occasional failures. Thus, for realistic use cases, leveraging redundancy-free distributed agreement during intervals with no failures improves performance significantly.

Many distributed systems require coordination between the components involved. With the steady growth of such systems, the probability of failures increases, which necessitates scalable fault-tolerant agreement protocols. The most common practical agreement protocol, for such scenarios, is leader-based atomic broadcast. In this work, we propose AllConcur, a distributed system that provides agreement through a leaderless concurrent atomic broadcast algorithm, thus, not suffering from the bottleneck of a central coordinator. In AllConcur, all components exchange messages concurrently through a logical overlay network that employs early termination to minimize the agreement latency. Our implementation of AllConcur supports standard sockets-based TCP as well as high-performance InfiniBand Verbs communications. AllConcur can handle up to 135 million requests per second and achieves 17x higher throughput than today's standard leader-based protocols, such as Libpaxos. Thus, AllConcur is highly competitive with regard to existing solutions and, due to its decentralized approach, enables hitherto unattainable system designs in a variety of fields.

This paper presents a programming model approach that explicitly addresses the programmability of scientific code by annotating imperative code with its algorithmic structural behavior. This information is used to create hierarchical structures, as opposed to the flat structure that most programming models work with, which allows sound code transformation at any level of the code, adjusting the granularity of parallelization simultaneously with other parameters to better exploit the available hardware resources.

Modern multicore systems offer huge computing potential. Exploiting large parallel systems is still a very challenging task, however, especially as many software developers still use overly-sequential programming models. refactoring tool support that allows the programmer to introduce and tune parallelism in an easy and effective way, exploiting high-level parallel patterns such as farms and pipelines. Using our approach, we achieve speedups of up to 21 on a 24-core shared-memory system for a number of realistic use-cases.

Hybrid parallel programming models combining distributed and shared memory paradigms are well established in high-performance computing. The classical prototype of hybrid programming in HPC is MPI/OpenMP, but many other combinations are being investigated. Recently, the data-dependency driven, task parallel model for shared memory parallelisation named StarSs has been suggested for usage in combination with MPI. In this paper we apply hybrid MPI/StarSs to a Lattice-Boltzmann code. In particular, we present the hybrid programming model, the benefits we expect, the challenges in porting, and finally a comparison of the performance of MPI/StarSs hybrid, MPI/OpenMP hybrid and the original MPI-only versions of the same code.

Spatial decomposition is a popular basis for parallelising code. Cast in the frame of task parallelism, calculations on a spatial domain can be treated as a task. If neighbouring domains interact and share results, access to the specific data needs to be synchronized to avoid race conditions. This is the case for a variety of applications, like most molecular dynamics and many computational fluid dynamics codes. Here we present an unexpected problem which can occur in dependency-driven task parallelization models like StarSs: the tasks accessing a specific spatial domain are treated as interdependent, as dependencies are detected automatically via memory addresses. Thus, the order in which tasks are generated will have a severe impact on the dependency tree. In the worst case, a complete serialization is reached and no two tasks can be calculated in parallel. We present the problem in detail based on an example from molecular dynamics, and introduce a theoretical framework to calculate the degree of serialization. Furthermore, we present strategies to avoid this unnecessary problem. We recommend treating these strategies as best practice when using dependency-driven task parallel programming models like StarSs on such scenarios.