Piccolo: Building Fast, Distributed Programs with Partitioned Tables

By Russell Power, Jinyang Li. USENIX Symposium on Operating Systems Design and Implementation, 2010

311 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/7526048dc4c5d8a06ef0679b6ab0ede3439add56

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Exploiting iterative-ness for parallel ML computations (2014, background): “Various stopping conditions may be used, such as when the objective function improvement slows sufﬁciently or just a given amount of time has been spent reﬁning.”
Petuum: A New Platform for Distributed Machine Learning on Big Data (2013, background): “Other systems provide low-level programming interfaces [16], [17], that, while powerful and versatile, do not yet offer higher-level general-purpose building blocks such as scheduling, model partitioning strategies, and managed communication that are key to simplifying the adoption of a wide range o”
CIEL: A Universal Execution Engine for Distributed Data-Flow Computing (2011, methodology): “…is especially noticeable in iterative computations, for which MapReduce Dryad Pregel Iterative MR Piccolo CIEL Feature [2] [18] [26] [28] [12] [21] [34] Dynamic control flow Task dependencies Fixed (2-stage) Fixed (DAG) Fixed (BSP) Fixed (2-stage) Fixed (1-stage) Dynamic Fault tolerance…”
Tachyon: Reliable, Memory Speed Storage for Cluster Computing Frameworks (2014, background): “Practitioners and researchers have built a wide array of programming frameworks [34– 36, 42, 53, 54] and storage systems [14, 15, 25, 26, 39] tailored to a variety of workloads.”
Efficient Memory Disaggregation with Infiniswap (2017, background): “However, existing proposals for memory disaggregation call for new architectures [11, 12, 49], new hardware designs [56, 57], and new programming models [63, 69], rendering them infeasible.”
Addressing the straggler problem for iterative convergent parallel ML (2016, methodology): “FlexRR uses the increasingly popular parameter server approach [3, 31, 35, 43] to storing and managing the shared state (i.”
Efficient Machine Learning for Big Data: A Review (2015, methodology): “There are several computing frameworks, e.g., Hadoop [56], SHadoop [57], ComMapReduce [58], Dryad [59], Pic-colo [60], and IBM parallel machine learning toolbox, such systems have the capabilities to scale up machine learning.”
Tupleware: Distributed Machine Learning on Small Clusters (2014, unknown): “Other programming models, such as FlumeJava [13], Ciel [27], and Piccolo [29] lack the low-level optimization potential that Tupleware’s algebra provides.”
One Trillion Edges: Graph Processing at Facebook-Scale (2015, background): “Piccolo [36] executes distributed graph computations on top of partitioned tables.”
Trends in big data analytics (2014, background): “Piccolo [72] allows distributed computations to storemutable state”
Solving the Straggler Problem with Bounded Staleness (2013, methodology): “LazyTables provides an API similar to the Piccolo [20] system.”
Distributed GraphLab: A Framework for Machine Learning in the Cloud (2012, background): “ta parallel abstractions to the iterative setting. However, these abstractions still do not support asynchronous computation. Bulk Synchronous Parallel (BSP) abstractions such as Pregel [25], Piccolo [33], and BPGL [16] do not naturally express asynchronicity. Onthe other hand, thesharedmemory Graph”
Distributed GraphLab: A Framework for Machine Learning in the Cloud (2012, background): “Surprisingly, many asynchronous abstractions like [32] do not ensure serializability or, like Piccolo [33], provide only basic mechanisms to recover from data races.”