Information aggregation is the process of summarizing information across the nodes of a distributed system. We present a hierarchical information aggregation system tailored for Peer-toPeer Grids which typically exhibit a high degree of volatility and heterogeneity of resources. Aggregation is performed in a scalable yet efficient way by merging data along the edges of a logical selfhealing tree with each inner node providing a summary view of the information delivered by the nodes of the corresponding subtree. We describe different tree management methods suitable for high-efficiency and high-scalability scenarios that take host capability and stability diversity into account to attenuate the impact of slow and/or unstable hosts. We propose an architecture covering all three phases of the aggregation process: Data gathering through a highly extensible sensing framework, data aggregation using reusable, fully isolated reduction networks, and applicationsensitive data delivery using a broad range of propagation strategies. Our solution combines the advantages of approaches based on Distributed Hash Tables (DHTs) (i.e., load balancing and self-maintenance) and hierarchical approaches (i.e., respecting administrative boundaries and resource limitations). Our approach is integrated into our Peer-to-Peer Grid platform Cohesion. We substantiate its effectiveness through performance measurements and demonstrate its applicability through a graphical monitoring solution leveraging our aggregation system.
Using the Barnes-Hut algorithm as an example we deal with the design of parallel algorithms that are able to exploit multicore CPUs and GPUs conjointly. Specifically, we demonstrate how to modularize a parallel application according to specific aspects of parallel execution. This allows for a flexible assignment of individual modules to the two parallel architectures based on their actual performance characteristics. Furthermore, we discuss a hybrid module for the most time consuming part of the algorithm that utilizes CPU and GPU simultaneously employing a novel load balancing heuristic. Our experimental evaluation shows that our method greatly increases overall efficiency by allowing to deploy the optimal configuration of modules for each individual computer system.Hannes Hannak is supported by a grant of the Landesgraduiertenförderung Baden-Württemberg.
Desktop Grids utilize the combined computing power of distributed resources to solve computationally hard problems. In contrast to conventional high-performance computing systems, this kind of parallel architecture exhibits a high degree of resource volatility and heterogeneity. Therefore, most existing projects in this area (e.g. BOINC) focus on trivial parallel algorithms, where no communication between nodes is necessary. However, there are non-trivial computational problems that cannot be efficiently solved by using such an approach but still are eligible for execution on Desktop Grids.In this paper we present a parallel formulation of the BarnesHut N-Body algorithm suitable for Desktop Grids, as a representative of this class. The redesign of this algorithm is based on our COHESION platform, which enables efficient peer-topeer communication on Desktop Grids. We describe how task generation and distribution is achieved and communication is minimized. In particular, we illustrate how a checkpointing and restart concept is used to make our parallel Barnes-Hut algorithm resilient to the unexpected withdrawal of peers. We finally present experimental evidence for the efficiency of our approach under various degrees of resource volatility.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.