In environments where exact synchronization between source data objects and cached copies is not achievable due to bandwidth or other resource constraints, stale (out-of-date) copies are permitted. It is desirable to minimize the overall divergence between source objects and cached copies by selectively refreshing modified objects. We call the online process of selecting which objects to refresh in order to minimize divergence best-effort synchronization. In most approaches to best-effort synchronization, the cache coordinates the process and selects objects to refresh. In this paper, we propose a best-effort synchronization scheduling policy that exploits cooperation between data sources and the cache. We also propose an implementation of our policy that incurs low communication overhead even in environments with very large numbers of sources. Our algorithm is adaptive to wide fluctuations in available resources and data update rates. Through experimental simulation over synthetic and real-world data, we demonstrate the effectiveness of our algorithm, and we quantify the significant decrease in divergence achievable with source cooperation.

The single source shortest path (SSSP) problem lacks parallel solutions which are fast and simultaneously work-efficient. We propose simple criteria which divide Dijkstra's sequential SSSP algorithm into a number of phases, such that the operations within a phase can be done in parallel. We give a PRAM algorithm based on these criteria and analyze its performance on random digraphs with random edge weights uniformly distributed in [0, 1]. We use

Abstract not found

We explain how the parallelization aspects of a large class of applications can be modeled as tree shaped computations. This model is particularly suited for NP-complete problems. One reason for this is that any computation on a nondeterministic machine can be emulated on a deterministic machine using a tree shaped computation. We then proceed to a particular example, the knapsack problem It turns out that a parallel depth first branch-and-bound algorithm based on tree shaped computations yields superlinear average speed-up using 1024 processors.